Simultaneous localization and mapping (SLAM) is considered as the core of building high-precision three-dimensional environmental maps. For this, we proposed a high-precision laser-based SLAM system. Aiming at the problem that the point cloud data obtained from laser is not representative and the iterative closest point method for calculating pose transformation is time-consuming, we proposed an efficient and stable matching algorithm. It uses fused feature points to align with the occupancy grid submaps with less registration error and is less time-consuming. Then, to address the problem in which the registration result falls into the local optimum early, we proposed a quadratic registration algorithm. This method effectively improves the initial value of the registration process. Finally, a time consistency and global consistency loop detection algorithm are used to reduce the cumulative error. The system we proposed has been tested on the University of Michigan North Campus Long-Term dataset and Cartographer dataset. Experiments show that our system has good accuracy under low-speed motion conditions (speeds ranging from 1 to 2  m  /  s).

This guest editorial introduces the Special Section on Advances in Remote Sensing for Forest Structure and Functions.

Sustainable forest management relies on the acquisition of timely (change detection) and accurate structural information of forest landscapes. Light detection and ranging (lidar) remote sensing platforms enable rapid, three-dimensional (3-D), structural data collection with a high spatial resolution. This study explores a functional carbon model applied to a dense, closed deciduous forest. Data are collected by manned airborne systems and unmanned aerial system, producing both lidar and structure-from-motion (SfM) 3-D mapping. A hybrid approach combining cost-effective SfM-generated data with lidar-derived digital elevation models also is explored, since the SfM fails to produce adequate terrain returns. Carbon modeling results are comparable to those achieved by the initial developers (r²  =  0.64 versus r²  =  0.72), despite the challenging uneven-aged forest environment. Vertical profiles, mapped utilizing a volumetric point density from the manned airborne lidar, are leveraged to train a binary classifier for disturbance detection. Producer’s accuracy, user’s accuracy, and Kappa statistic for disturbance detection are 94.1%, 92.2%, and 89.8%, respectively, showing a high likelihood of detecting disturbances (harvesting). The results bode well for the use of unmanned aerial system (UAS) systems, and either lidar or SfM, to assess forest stocking. Although disturbance detection is successful, further study is required to validate the use of UAS, and especially SfM, for this task.

Forest canopy height is a very important forest structural attribute. LiDAR and SAR are able to penetrate the forest canopy to obtain information on the understory and canopy vertical structure. But the single data of LiDAR or SAR has its own shortcomings in forest height extraction. We jointly use LiDAR and ALOS PALSAR data to retrieve forest canopy height. First, the extinction degree of the canopy is extracted using airborne LiDAR. The canopy is assumed to be uniform, and the extinction degree is divided by the canopy height to obtain the average extinction coefficient. Then, the extinction coefficient is substituted into random volume over ground (RVoG), and the forest canopy height is obtained. Experimental results showed that the collaborative inversion algorithm based on RVoG model proposed in this paper improves the accuracy of forest canopy height retrieval.

To better assess spatial variabilities in subtropical secondary forest biomass, the goal of the present study was to employ small-footprint, discrete-return light detection and ranging and unmanned aerial vehicles, integrated with structure from motion (UAV-SFM) data, to accurately estimate the stand characteristic parameters of subtropical forests. The Lorey’s height, density, basal area, volume, and aboveground biomass (AGB) of 30 plots (Ø10  m) were modeled using an array of SFM point cloud metrics and image color metrics (forestry vegetation coverage and color index set). Further, the individual models developed using stepwise multiregression analysis and a branch-and-bound algorithm were employed to examine the best models. Overall, the results indicated that the coefficient of determination (R²) for Lorey’s height (R²  =  0.58 to 0.95), volume (R²  =  0.29 to 0.71), and AGB (R²  =  0.27 to 0.64) were significantly enhanced compared with the density (R²  =  0.18 to 0.48), and basal area (R²  =  0.24 to 0.52). Utilizing independent stand-level data from ground inventory, our results revealed that, overall, the model fit was significant for most stand characteristic parameters, with relationships close to a 1:1 line, thereby indicating no significant bias [relative root mean square error   (  rRMSE  )    =  5.32  %   to 27.04%]. Our image metrics had a statistically significant association with Lorey’s height (rRMSE decreased by 0.08%), volume (rRMSE decreased by 0.58%), and AGB (rRMSE decreased by 3.11%), which provided additional explanatory power in the regression analysis. This research demonstrated the potential of UAV-SFM as a technology to accurately assess subtropical forest carbon while providing an improved elucidation of the positive effects of image color metrics.

Traditional field measurement methods are usually time-consuming and costly. With recent developments in remote sensing, precise estimation of some key forest parameters is becoming a close reality. One of the robust technologies being utilized for this aim is light detection and ranging (LiDAR). We used LiDAR to derive tree height and canopy density metrics using pixel values, texture features, and vegetation indices obtained from WorldView-3 imagery to estimate volume per hectare, mean height, dominant height, and number of trees per hectare. A total of 58 sample plots, predominantly composed of Pinus brutia, which is one of the most abundant tree species in Turkey, were analyzed. The adjusted R² values obtained by the best regression models, using LiDAR-derived metrics alone, were 0.66, 0.73, 0.83, and 0.83 for the volume per hectare, number of trees per hectare, and average and dominant heights, respectively. After integrating LiDAR-derived metrics with WorldView-3 imagery band values and subsequently with vegetation indices, higher adjusted R² values of 0.70 and 0.77, respectively, were obtained for the volume per hectare. In contrast, incorporating texture features besides other parameters had no positive effect on the accuracy of estimation.

Sun-induced chlorophyll fluorescence (SIF), now observed from space on a global scale, has been shown to be a powerful proxy for photosynthetic activity. Long-term in situ field canopy SIF measurements are improving, enabling better understanding of SIF signal and support to satellite missions. Since SIF retrievals rely on the absolute irradiance radiance measurements, accurate outdoor radiometric calibration is necessary to provide accurate radiance data, especially for long-term continuous measurements on forest stands. However, standard laboratory calibration methods are not practical to be performed for outdoor long-term measurements, and several in-field alternative methods are generally used, which may result in some uncertainties. Here we evaluate the effects of different radiometric calibration methods on SIF retrievals and its relation with gross primary productivity (GPP) at the canopy level. Three widely used methods are used, i.e., well-established laboratory methods with integrating sphere, laboratory-calibrated reference spectrometer (ASD FieldSpec Pro), and a light calibration source with a white reference panel. Our results indicate that different radiometric calibrations have marginal effects on vegetation indices but have significant effects on the SIF absolute value, and have slight effects on diurnal and seasonal patterns of SIF. Moreover, the relationships between GPP and SIF retrieved using different calibration coefficients are similar at both diurnal and seasonal scales, but with different regression slopes. Therefore, we recommend that a standard laboratory radiometric calibration is conducted using an integrating sphere before installation in the field and a regular in-field calibration with a well-calibrated spectrometer and white reference, especially for long-term observation above a forest canopy. Our findings have strong implications for the ongoing and future ground canopy SIF measurements and suggest the need for a consistent radiometric calibration method, especially for cross-site comparison studies.

The continuous forest inventory is critical to forestry monitoring, Earth science applications, and global carbon cycle researching. Traditional field measurements of forest inventory are time-consuming. Targeting the problem, our study proposes an approach for monitoring the forest tree heights by combining geoscience laser altimeter system data with land surface reflectance data from Landsat TM based on artificial neural network (ANN) model in the Nanning research region. The land surface reflectance data are used as inputs of ANN model. The GLAS data are used to derive tree heights for training ANN model. Then the forest height spatial distribution and corresponding height change during 2003 to 2006 are investigated. The validation shows the high precision of modeling tree heights by, respectively, calculating determination coefficient (R²) and root-mean-square error (RMSE) of evaluations from 2003 to 2006 (R²  =  0.667, mean RMSE  =  3.219  m). Finally, the change trends of forest tree heights in Nanning region are obtained by calculating the difference of tree heights from 2003 to 2006.

Coarse spatial resolution (CSR) time series have been successfully used at regional scale to produce homogeneous and up-to-date forest cover maps. This study aims to classify CSR time series using a nomenclature as detailed as national forest inventories nomenclatures. To identify best practices for classifying time series, three algorithms are compared: maximum likelihood, support vector machine, and random forest. For each algorithm, training, temporal compositing, and selection of input features have been optimized. Results establish a clear improvement in classification accuracy when red, near-infrared, and short-wave infrared spectral bands are used instead of vegetation indices. Temporal compositing has a major impact when the whole phenological cycle is used for 3 consecutive years. Random forest produces the best classification, support vector machine appears to be sensitive to overtraining, and maximum likelihood is unable to deal with the complex characteristics of forest and natural vegetation classes.

The accurate estimation and inversion of stand characteristic parameters and biomass is the premise for the evaluation of forest productivity, forest ecosystem carbon storage, and ecological forest service capacity. The horizontal accuracies of structure from motion (SfM) and light detection and ranging (LiDAR) point clouds were identical, whereas the descriptions of the upper canopy indices were similar (R  =  0.91 to 0.93). However, the descriptions of the lower canopy indices and vertical structure variables were different based on comparisons of the two point cloud types in the same region. Variables that described the vertical structure of the canopy in the SfM point clouds were much less important in the factor estimation models, exhibiting high correlations with basal area and volume. The factor estimation models with high correlations with tree height and diameter at breast height (DBH) exhibited similar importance for LiDAR variables. LiDAR (R²  =  0.515 to 0.895) was superior to SfM (R²  =  0.63 to 0.835) in estimating the same parameters, particularly those with high correlations with vertical structures. The combination of unmanned aerial vehicle aerial photogrammetry and LiDAR technology might enable the estimation of stand characteristic parameters and the inversion of biomass, as well as the estimation of characteristic parameters such as Lorey’s H and DBH, which are highly sensitive for large datasets. The LiDAR and SfM model outcomes were comparable. We found that the model outcomes exhibited the greatest differences for volume (ΔrRMSE  %    =  13.94  %  ) and the lowest for average DBH (ΔrRMSE  =  0.8  %  )

The individual tree crown (ITC) segmentation algorithm based on aerial images is a prerequisite for understanding tree growth, tree species competition, and biomass assessment. We combine superpixel segmentation and topological graph methods to separate the ITC effectively from aerial images. First, the aerial images of forest plots captured by drones was segmented by simple linear iterative clustering of superpixel algorithm, and the crown boundaries of aerial images were obtained by deep learning concept of holistically nested edge detection (HED) network. Second, the similarity weights of neighboring superpixels were measured by three indices, i.e., the difference in color value, the number of intersecting pixels, and the number of boundary pixels defined by HED network in the intersecting area. Finally, the minimum spanning tree topological method was adopted to generate the connected trees of aerial images at the superpixel scale, and the superpixels were merged to realize ITC segmentation depending on the calculated similarity weights. This method was tested on the aerial images of three forest plots with different stand structural features, and the accuracies of the algorithm were evaluated by comparing the results of our algorithm with field measurements. Mixed growth of the withered trees and healthy trees is in the forest plot 1, which complicates the ITC segmentation process and only achieves 86% accuracy. The forest plot 2 with same tree species and approximately sized tree crowns obtains the highest ITC accuracy of 92%. The forest plot 3 has various sizes of tree crowns and is influenced by the upper-right solar illumination, which increases the difficulty of ITC segmentation using our algorithm and obtains 87% accuracy. Overall, the method proposed has promising potential for ITC segmentation from forest aerial images, which provides a new concept based on image processing technique suitable for various types of forests.

The easy and ready access to Landsat datasets and the ever-lowering costs of computing make it feasible to monitor the Earth’s land cover at Landsat resolutions of 30 m. However, producing forest-cover products rapidly and on a large scale, such as intercontinental or global, is still a challenging task. By utilizing the huge catalog of satellite imagery as well as the high-performance computing capacity of Google Earth Engine, we proposed an automated pipeline for generating 30-m resolution global-scale forest map from time-series of Landsat images. We describe the methods to create products of forest cover at a global scale. First, we partitioned the landscapes into subregions of similar forest type and spatial continuity. Then, a multisource forest/nonforest sample set was established for machine algorithm learning training. Finally, a random forest classifier algorithm was used to obtain samples automatically, extract the characteristics of satellite images, and establish the forest/nonforest classifier models. Taking Landsat8 images in 2018 as a case, a novel 30-m resolution global forest cover (GFC30) map has been produced. The result shows that by the end of 2018, the total forest area in the world was 3.71  ×  10⁹  ha. The accuracy evaluation of GFC30 for 2018 was carried out using verification points via stratified random sampling of a MODIS land cover map (MCD12C1 product in 2012) and verified on high-resolution satellite imagery (e.g., Google Earth). According to the validation result, the overall accuracy of GFC30 for 2018 is 90.94%.

The accurate retrieval of forest functional and structural parameters is of great significance in the scientific research of ecosystem, global change, and carbon and nitrogen cycles. Recently, an unmanned aerial vehicle (UAV) hyperspectral imaging system provides a cost-effective way to capture the hyperspectral imageries from any points of the hemisphere above a forest canopy. However, compared with single-angle hyperspectral images, the multiangle hyperspectral images provide more information about forest functional and structural characteristics. We developed a semiautomatic multiangle observation method using a UAV hyperspectral imaging system and successfully collected the multiangle hyperspectral imageries with clear hotspots of broadleaf and coniferous forest canopies. Our results indicated that the hotspot of a forest canopy had a great effect on the reflectance, normalized difference vegetation index (NDVI), and enhanced vegetation index (EVI) of forests. The maximum values of canopy reflectance and EVI were found at the hotspot position, while the minimum NDVI was at the hotspot. Moreover, the hotspot effect was similar in both broadleaf and coniferous forests. Although the hotspot had no obvious effects on the photochemical reflectance index (PRI), different view zenith angles had a great effect on PRI. Our findings provide a solid foundation for retrieving forest structural parameters using fully automatic multiangle hyperspectral imaging system at both aerial and satellite platforms.

Forest has complex horizontal and vertical structures, so it is difficult to accurately identify tree species from remote sensing images only by spectral characteristics. Based on the airborne hyperspectral and LiDAR data sets, adding spatial features into tree species identifying procedure was studied in Gaofeng Forest Farm of Guangxi, China. As a result, a tree species classification method of synergizing spectral, spatial, and vertical structure features was proposed. First, principal components (PCs) were extracted from the hyperspectral data acquired by the AISA Eagle II sensor as spectral features. Four typical spatial features, gray level co-occurrence matrix, extended morphological profile, extended multiattribute profile, and multiscale guided filtering (MGF), were derived from the principal component bands. Meanwhile, digital elevation model, digital surface model (DSM), and normalized DSM were calculated from the LiDAR data to represent the characteristics of vertical structure. Then 13 spectral–spatial–vertical feature combinations were built and compared to classify the tree species using random forest classifier. The experimental results show that both spatial and vertical structure features can effectively improve the classification accuracy of tree species when banding with spectral features, and the feature combinations including multiscale spatial features is more conducive to improve classification accuracy. Compared with the method of using spectral features only, the methods with spatial features or vertical structure features could enhance the overall accuracy (OA) maximally by 49.49% and 14.65%. Among all the feature combinations, the synergy of PC bands, DSM, and MGF features achieves the highest classification accuracy, with OA 96.10% and kappa 0.9575. The spatial features have a great application potential for tree species classification in a plantation forest environment.

Stand density is one of the important forest structure parameters. QuickBird and Worldview-2 high spatial resolution remote sensing images were compared. The Jiangle state-owned forest farm in Fujian Province was the experimental area. A spectral local maximum filtering method was used to extract the number of individual trees in a mountain forest. Nonlinear quadratic polynomial regression models were established for the number of local maximum points and the actual stand density. The near-infrared II waveband of Worldview-2 imagery was used to extract the pure Chinese fir stands with an accuracy of 72.5%, R²  =  0.502, and RMSE  =  35.77 and the Masson pine stands with an accuracy of 78.35%, R²  =  0.754, RMSE  =  41.46, while the accuracy of all unclassified stands was only 0.2907. The results showed that the classification of tree species can improve the accuracy of modeling. The quadratic polynomial model using Worldview-2 image data achieved better results. Stand density in the Jiangle state-owned forest farm was extracted using the NIR II band after fusion of Worldview-2. A stand density planning map was constructed using the best models applied to the different forest types. The method of combining high-spatial resolution imagery and sampling plots to estimate stand density was proven to be feasible.

Hyperspectral image (HSI) classification is one of the significant research topics in the remote sensing community. The high dimensionality of the hyperspectral data, the high correlation among pixels, and the availability of fewer numbers of training samples affect the HSI classification accuracy. We propose an approach to extract the best representative bands from the high-dimensional imagery for better classification. Initially, the spectral bands are extracted by re-representing the traditional principal component analysis in terms of Hebbian learning, formulated and solved as a fuzzy optimization problem. Next, a spatial filter is applied to these spectral bands to obtain the smoothed image that preserves the spatial details. Finally, the spectral and spatial features are trained with the nonlinear support vector machine with the radial basis function kernel to obtain the classification map. Performance of the proposed approach is tested by varying different values of the parameters used in our model. The classification accuracy of the proposed approach is compared with the state-of-the-art techniques, which proves the effectiveness of the proposed methodology. The proposed approach can be applied in real-world applications, such as food quality, environment change detection, mineralogy, and pharmaceutical drug design.

Wetlands are among the most valuable natural resources, being highly beneficial to both the environment and humans. Therefore, it is very important to map and monitor wetlands. Although various remote sensing datasets, including optical, synthetic aperture radar (SAR), light detection and ranging (LiDAR) imagery, have been widely applied to classify wetlands, it is still required to discuss the advantages/limitations of each of these datasets and suggest the best remote sensing methodology for wetland mapping. Thus, the Terra Nova National Park, located in Newfoundland, Canada, was initially selected as the study area to develop a supervised classification method along with object-based image analysis. To this end, different remote sensing-based scenarios were investigated using individual optical, SAR, and LiDAR datasets, as well as their various combinations. In addition, for achieving the highest accuracy, the effects of segmentation scales and the tuning parameters of the random forest (RF) classifier were examined. The results showed that a combination of optical, SAR, and LiDAR images with the segmentation scale of 150, the RF depth of 20, and the RF minimum sample number of 5 provided the highest classification accuracy with the overall accuracy of 87.2%. Moreover, based on the results, approximately 21% and 79% of the study area are covered by wetlands and nonwetlands, respectively. The proposed methodology shows an optimum scenario for future wetland classification tasks and can assist stakeholders in the effective management of wetlands and establishment of necessary policies.

Accurate remote sensing retrieval of chlorophyll-α (Chl-α) concentrations in inland waters raises a challenge due to the optical complexity of water constituents. Five Chl-α retrieval models, including single-band, band ratio, three-band, four-band, and partial least square models, were established with the measured spectra and Chl-α concentrations were measured at 36 stations in Panjiakou and Daheiting Reservoirs. To improve the Chl-α retrieval accuracy, three ensemble models, namely, entropy weight-based ensemble model (EW-EM), set pair analysis-based ensemble model (SPA-EM), and Bayesian model averaging-based ensemble model (BMA-EM), were developed for Chl-α retrieval with the weighted average of the five Chl-α retrieval models. All models were evaluated based on random calibration and validation samples. Ensemble modeling improved the Chl-α retrieval accuracy through integrating multiple Chl-α retrieval models. Compared to EW-EM and SPA-EM, BMA-EM could not only improve the Chl-α retrieval accuracy but also provide reliable confidence intervals for Chl-α retrieval. Ensemble modeling has application prospects in remote sensing retrieval of water constituents in inland waters.

Pavement condition monitoring is fundamental for the efficient allocation of resources in transportation asset management. However, data collection involves laborious and costly procedures. Our study intends to investigate the usage of remote sensing data for network-level pavement condition assessment that offers a more cost-effective alternative and a rapid infrastructure assessment tool that can be used in the aftermath of natural disasters. Based on an extensive literature review, a data mining framework was established to train models that predict the pavement condition of different road segments. The framework exploits the inherent information of multispectral images by generating spectral related attributes. To identify pavement sampling areas, an automated procedure using image segmentation replaces manual surface digitizing. Unlike previous research, different classification models were used to approximate the mapping function from spectral information to pavement conditions. A preliminary case study was conducted with data provided by the City of Dallas and multispectral images acquired from the Texas Natural Resources Information System. The mean-shift segmentation algorithm was used to locate noise introducing areas on the pavement surface. Four different classification models were trained using k-nearest neighbors, naïve Bayes, support vector machines, and a multilayer perceptron. The developed models were employed to predict the road surface condition class of a test set not included in the training procedure. The multilayer perceptron presented the highest accuracy level of 71%, showing that the framework might have the potential for future implementation.

Reducing emissions from deforestation and forest degradation (REDD+) has emerged as a global climate change mitigation initiative under negotiation by the United Nations Framework Convention on Climate Change aimed at providing financial support to the developing countries for conserving respective forests. To implement the REDD+ initiative, developing countries need to estimate, among other necessitates, the activity data (i.e., pattern and process) of respective deforestation and forest degradation. Bangladesh is steadily progressing through its REDD+ roadmap. However, an important research issue to address includes using remote sensing technology to detect activity data for deforestation in a spatially explicit manner following the recommended good practice guidelines by the Intergovernmental Panel on Climate Change. This study mapped the activity data for deforestation of a mixed forest in Bangladesh during 1995 to 2015, applying Monte-Carlo spectral unmixing classification algorithm to Landsat images in CLASlite software. The classification was verified using independently drawn reference points from high-resolution Google Earth images. A postclassification comparison method was applied to generate landcover transition matrices. The outputs were highly accurate maps (overall accuracy >90  %  ) and statics of activity data for deforestation of the study area. The approaches and findings may have significant implications in adopting any REDD+ project in Bangladesh.

In order to extract urban impervious surfaces (ISs) accurately, optical and synthetic aperture radar (SAR) images fusion was recognized as one promising method. However, most fusion methods currently focus on feature-level and decision-level fusions. There are only a few studies exploring the performance of the fused image at the pixel level for IS extraction. Therefore, we introduced the shearlet transform to fuse Landsat-8 and TerraSAR images and evaluated the fused image by comparing it to those obtained using conventional image fusion methods. The IS from the fused images using the support vector machine algorithm is extracted and compared. Experimental results indicate some interesting findings. First, the shearlet transform can fully retain the spectral information from the optical image and the spatial information from the SAR image. Second, the IS extraction from the fused image with the shearlet transform achieved the highest accuracy with an overall accuracy of 95.1% and a Kappa coefficient of 0.8792, which confirmed the proposed method is applicable to IS extraction. We can conclude that an effective pixel-level fusion algorithm for optical and SAR images can significantly improve the extraction accuracy of urban IS. Our research could provide an innovative fusion technique and also could serve as a meaningful reference for further applications of optical and SAR imagery. In addition, the potential of SAR data in IS extraction should be further investigated.

Image navigation is a primary process for on-orbit optical payloads involving environmental disaster monitoring, meteorological observation, and the positioning and tracking of space-aeronautics targets. However, because they are affected by solar illumination and orbital heat flux, as well as shock and vibration during launch, the installation structures between the instruments and satellite platform, especially for geostationary satellites, will inevitably generate a displacement resulting in the reduction of positioning accuracy. During the application of FengYun-4A (FY-4A), it is found that the further away from the subsatellite point, the greater the positioning error of Advanced Geostationary Radiation Imager (AGRI) on FY-4A will be. The positioning error can reach 14 pixels at the Arabian Peninsula in operational images. In addition, compared with orbital and attitude measurement errors, long-term observations show that the installation matrix is likely to be the most significant factor determining the navigation accuracy of AGRI. Therefore, an on-orbit installation matrix calibration approach as well as a high-precision navigation algorithm is proposed to modify the positioning error of AGRI. Experimental results show that the navigation error of the processed images corrected by the proposed method can be reduced to 1.3 pixels, which greatly improves the navigation procession of AGRI. In general, this method could be a supplement to the correction of positioning error for geostationary payloads.

The difficulty of accurate and large-scale measurement for surface parameters limits the regional surface soil moisture (SSM) estimation using synthetic aperture radar (SAR). Moreover, the coarse resolution of soil moisture products generated by existing methods, which fuse SAR and passive microwave products, cannot fully satisfy the requirement of specific regional applications. To solve this problem, an SAR-optical data fusion method for soil moisture estimation (SOFSME) based on a cascade neural network is proposed in this study. SOFSME obtains surface parameters from historical soil moisture images and related environmental images to estimate a SSM image with high resolution at large scale from Sentinel-1A C-band SAR data. Validation experiments in single and multiple land-use type areas showed that the SOFSME performed best on bare soil areas with a median root mean square error of 0.0203. The median universal image quality index of estimated soil moisture image was 0.1454, which was better for single cropland areas than multi-land-use type areas. The Pearson correlation coefficient showed a median value of 0.7645 in both experiments. These results showed that the SOFSME had high accuracy, availability, and stability in regional soil moisture estimation. Compared with existing methods, the SOFSME can provide high-quality soil moisture images and does not directly depend on field measurement data. Thus, the proposed SOFSME method is of great value for high-resolution soil moisture estimation in more regional applications.

The forest structural attributes are required information for sustainable forest management. The use of different remote sensing sources has been investigated intensively as a new potential and an alternative for the forest stand characteristics estimation during the last few years. This research purpose was to examine the phased array type L-band synthetic aperture radar (PALSAR) onboard the Advanced Land Observing Satellite (ALOS) data ability in order to estimate stand volume, basal area, and tree density in the Hyrcanian forests of Iran with high composition and structure variations. The required preprocessing and processing steps were performed on the ALOS/PALSAR raw data, and the corresponding values of circular plots were extracted on all SAR data. The modeling of forest structure attributes was performed using field-collected attributes by the k-nearest neighbor (kNN), support vector machine (SVM), artificial neural network (ANN), and multiple linear regression (MLR) algorithms. The modeling validity was performed by unemployed plots and by the absolute and relative root mean square error (RMSE) and bias measures. The results of this study have shown that although the results of ANN, SVM, and kNN algorithm were not very different but compared to MLR algorithm, they had better performance. In addition, based on the results of this study, the ANN algorithm showed slightly better performance in forest attribute prediction than the other used algorithms. The results were 34.56%, 27.65%, and 31.16% in relative RMSE for stem volume, basal area, and tree density prediction.

Recent reports from Isfahan, Iran, have announced the possibility of potential land subsidence due to the persistence of drought, interbasin water transfer, and groundwater overexploitation. By this token, we seek to employ the persistent scatterer interferometric synthetic aperture radar technique to detect and measure the land subsidence phenomenon in the Isfahan metropolitan area. The Sentinel interferometric synthetic aperture radar dataset was used for detecting land subsidence phenomena. The corresponding data pairs were acquired and analyzed for the period from 2014 to 2019. Our prime objective is to map the spatial and temporal variations in surface deformation in Isfahan using the multitemporal persistent scatterer interferometric synthetic aperture radar (MT-PSI) technique. Results indicate that Isfahan had been undergoing subsidence throughout the observation period of 2014 to 2019 at an estimated rate of −5 to −100  mm  /  year. Subsidence has increased from south to north of metropolitan areas from 2014 to 2019. The MT-PSI time series during the observation period correspond well with fluctuations in the level of the groundwater. The subsidence in the Isfahan region is mainly due to human activities. The estimated subsidence rates in the city for the period from 2014 to 2019 showed no signs of high subsidence rate in the cultural-historic and economic sites of the urban regions of Isfahan.

Spectral unmixing (SU) has been a subject of particular attention in the hyperspectral imaging literature. Most SU algorithms are based on the linear mixing model (LMM), which assumes that one pixel of the image is the linear combination of a given number of pure spectra called endmembers, weighted by their coefficients called abundances. SU is a technique to identify these endmembers and their relative abundances. We present an LMM approach based on nonnegative matrix factorization, combining the minimum volume constraint (MVC) and Kullback–Leibler (KL) divergence referred to as KL-MVC. The proposed method is evaluated using synthetic images with different noise levels and real images with different methods of initialization, and high performance has been achieved compared with the widely used LMM-based methods.

Monitoring crops at a fine scale is critical because it provides information crucial for assessing the influence of increased food production on sustainable management of agricultural landscapes. We assessed the potential of synthetic aperture radar (SAR) Sentinel-1 (S-1) and optical Sentinel-2 (S-2) images to derive crop biophysical parameters of wheat and rapeseed in Brittany, France. We generated a dataset of 29 features, including spectral bands and vegetation indices derived from S-2 images, and backscattering coefficients and polarimetric indicators derived from S-1 images. Then, we compared the respective value of S-1 and S-2 features to estimate crop LAI, biomass, and water content (WC) using a Gaussian processes regression. The results show that best S-2-based models were achieved using the green band, NIR bands, and vegetation indices for the wheat WC, LAI, and biomass, respectively, and the shortwave-infrared bands for rapeseed biomass. Concerning S-1-based models, the σ⁰VH  :  σ⁰VV ratio was the most relevant feature for wheat LAI and rapeseed biomass, and the Shannon entropy polarization contribution best performed for wheat WC. We highlighted not only the value of optical S-2 images but also the importance of polarimetric indicators derived from SAR S-1 images for estimating crop biophysical parameters.

Detailed building façade structures are significant parts of three-dimensional building models. A photogrammetric point cloud, generated from oblique unmanned aerial system imagery, is a type of data applied in building reconstruction. Its positive characteristics, alongside its challenging qualities, have provoked discussions related to our study. To obtain highly accurate window detection results, we propose a regular window detection method using a global regularity level set. Our method first detects the window boundaries from point clouds using a hole-based boundary extraction method. The minimum bounding rectangle of each boundary point cloud is then used specifically to represent the window. Next, façade partitioning is applied to subdivide the rectified façade images into a series of slices followed by cells. Each cell contains only one window rectangle after the processing. Finally, a global regularity level set algorithm is developed to optimize rectangular windows in horizontal and vertical slices using façade images. The proposed method was validated through a comparison with the slicing method on data from two façades. The accuracy of the window sizes in height and width was also discussed. The experiment results indicate that our method can obtain regular window locations and shapes from photogrammetric point clouds.

Our study uses artificial intelligence and machine learning methods for dimensionality reduction, feature extraction, and classification of hyperspectral remote sensing images to detect subtle differences between different spectra. Our study also used hyperspectral remote sensing image data gathered by airborne visible infrared imaging spectrometer over the Indian Pines and Salinas test sites. Based on experimental analysis and comparison of mainstream classifiers and advanced algorithms, we propose a new classification mode for the hyperspectral image based on a two-channel fusion of convolutional neural network-support vector machine (CNN-SVM). This model gives full play to the advantages of CNN in feature extraction and the advantages of SVM in the classification. It also maximizes the generalization and accuracy of classification. The CNN is based on two-channel feature extraction, including spatial features and spectral features, which utilizes the characteristics of hyperspectral remote sensing images for classification. Experimental results show that the research is both theoretically and practically significant for improving the quality and accuracy of identification and classification.

An increase in forest damage caused by the European spruce bark beetle (Ips typographus L.) has recently been observed in many Central European forests such as the forests of the Central German Uplands. For these regions, which are highly vulnerable due to their tree species composition and an increasing frequency of extreme weather events, the availability of time- and cost-efficient monitoring procedures is of particular importance. Thus it was the objective of our study to evaluate the potential of Sentinel-2 data for large-area detection of bark beetle infestation. Using change detection, infestation maps were produced for the Saxon Switzerland National Park Forest District (SAXS) and the Black Forest National Park (BFOR). An accuracy assessment provided a user’s accuracy of 97% (SAXS) and 88% (BFOR), respectively. The calculation of the producer’s accuracy (AP) resulted in 34% (SAXS) and 4% (BFOR). However, when small infestation areas were excluded from stereoscopic reference data, AP increased up to 52% (SAXS) and 40% (BFOR), respectively. Taking these findings into account, our study highlights both the high value of Sentinel-2-based change detection for area-wide bark beetle monitoring and its limitation to large contiguous infestation areas.

In order to overcome errors caused by season, shooting angle, and other factors in multitemporal remote sensing image change detection, a method based on old-temporal vector data and new-temporal image data is proposed under the premise that changes of images are normally less than the unchanged ones. Getting the object through incremental segmentation under the constraints of the previous vector data, we extract its textural and spectral features to get the dataset by the transform of principal component analysis. After this, the isolation forest method is used to calculate the object’s change index, and the change threshold is obtained by the Bayes method. We conduct two experiments. The effectiveness of the proposed method was verified by comparing image–image and vector–image change detection methods as well as Mahalanobis distance and isolation forest change methods for which the accuracy rate of experiment 1 is 92.35% and that of experiment 2 is 93.18%.

Increased demand for grazing resources has prompted grassland productivity optimization through fertilization. Despite these initiatives, there is no comprehensive framework for monitoring productivity dynamics in fertilized grasslands. In this regard, we evaluated the potential of field-based hyperspectral data in characterizing foliar chlorophyll content of grass grown under complex fertilizer treatments. Data analysis was done using advanced regression methodologies. Our study showed that chlorophyll content significantly varies among grasses treated with different fertilizer combinations. Further, foliar chlorophyll content estimation results can be accurately derived from the combined use of hyperspectral multiband and vegetation indices. High accuracies were attained as indicated by the mean of squared residuals of 5.41  μg m^−  2, 90.72% of explained variance, root-mean-square error of 4.02  μg m^−  2, and r² of 0.91. In addition, the variable importance modeling depicted sR 435/835 nm; nDVI 415/735, nDVI 545/895, 720 nm, nDVI 515/ 835, and 800 nm as the key foliar chlorophyll predictor variables for the grasses fertilized with 11 combinations of ammonium nitrate and ammonium sulfate combined with lime and phosphorus, as well as a control. These findings underscore the utility of spectroscopic proximal data for the provision of inherent subtle grass characteristics and location-specific information required to inform optimal grassland management strategies.

The goal of this research is to develop a general deep learning solution for atmospheric correction and target detection using multiple hyperspectral scenes. It is assumed that the scenes differ only in range and viewing angles, that they are acquired in rapid sequence using an airborne sensor orbiting a target, and that the target and the atmosphere remain invariant within the time scale of the collection. Several hundred thousand hyperspectral simulations were performed using the MODTRAN model and were used to train the deep learning solution, as well as to validate the proposed method. The input to the deep learning solution is a matrix of the simulated radiances at the sensor as function of wavelength and elevation angles. The output is atmospheric upwelling, downwelling, and transmission. This solution is repeated for all or a subset of pixels in the scene. We focus on emissive properties of targets, and simulations are performed in the longwave infrared between 7.5 and 12  μm. Results show that the proposed method is computationally efficient and it can characterize the atmosphere and retrieve the target spectral emissivity within one order of magnitude errors or less when compared with the original MODTRAN simulations.

Farmers and growers typically use approaches based on the crop environment and local meteorology, many of which are labor-intensive, to predict crop yield. These approaches have found broad acceptance but lack real-time and physiological feedback for near-daily management purposes. This is true for broad-acre crops, such as snap bean, which is valued at hundreds of millions of dollars in the annual agricultural market. We aim to investigate the relationships between snap bean yield and plant spectral and biophysical information, collected using a hyperspectral spectroradiometer (400 to 2500 nm). The experiment focused on 48 single snap bean plants (cv. Huntington) in a controlled greenhouse environment during the growth period (69 days). We used applicable accuracy and precision metrics from partial least squares regression and cross-validation methods to evaluate the predictive ability of two harvest stages, namely an early-harvest and late-harvest stage, against our yield indicator (bean pod weight). Four different spectral data sets were used to investigate whether such oversampled, hyperspectral data sets could accurately and precisely model observed variability in yield, in terms of the coefficient of determination (R²) and root-mean-square error (RMSE). The objective of our approach hinges on the philosophy that selected spectral bands from this study, i.e., those that best explain yield variability, can be downsampled from a hyperspectral system for use in a more cost-effective, operational multispectral sensor. Our results suggested the optimal period for spectral evaluation of snap bean yield is 20 to 25 or 32 days prior to harvest for the early- and late-harvest stages, respectively, with the best model performing at a low RMSE (3.02  g plant^−  1) and a high coefficient of determination (R²  =  0.72). An unmanned aerial systems-mounted, affordable, and wavelength-programmable multispectral imager, with bands corresponding to those identified, could provide a near real-time and reliable yield estimate prior to harvest.

In civil engineering, determining the subsurface moisture content and the water-to-cement (w  /  c) ratio of concrete structures in the field is a challenging task. Moisture content of concrete structures regards the cement hydration level and short-term strength development of concrete, as well as the phase-distribution (solid, gaseous, and liquid) inside concrete and its long-term durability. We describe a microwave nondestructive evaluation technique using noncontact/nonintrusive synthetic aperture radar (SAR) imaging to estimate both the w  /  c ratio and subsurface moisture content of concrete panels. Concrete panels made of three w  /  c ratios (0.4, 0.5, and 0.55) were air dried for ∼3 months and scanned by a laboratory 10-GHz SAR imaging system. From this research, we found that SAR imaging can capture the moisture content variation and its spatial distribution. A parameter of SAR images, the critical contour area, is identified and proposed for estimating subsurface moisture content and the w  /  c ratio of concrete panels. At a given w  /  c ratio, the critical contour area increases with the increase of moisture content. At a given moisture content, the critical contour area decreases with the increase of the w  /  c ratio.

A method was proposed for correcting the attenuation of C-band weather radar reflectivity data based on a 2-D video disdrometer. The relationship between the C-band weather radar reflectivity attenuation rate k and radar reflectivity factor Z was derived using radar meteorological equations. A video disdrometer was used to correct the deformation of large (1 to 9 mm) raindrops, and the coefficients a and b of the k  −  Z relationships for different precipitation types under Rayleigh scattering and Mie scattering conditions were inverted separately. Data from a weather detection instrument and rain gauge were combined with Newton’s iteration method for database-by-database accumulative corrections of the C-band radar reflectivity factor. Observed C-band weather radar data recorded in Xichang city and video disdrometer data from Xichang, Dechang, Mianning, and Xide counties during 2016 and 2017 were selected (all areas are in Liangshan Prefecture, Sichuan Province). The relationships between the raindrop size distribution and the radar reflectivity factor and attenuation rate in stratiform and convective cloud precipitation were analyzed, and the k  −  Z relationships under different precipitation types in Liangshan were presented. Additionally, radar reflectivity attenuation correction errors were analyzed using observed C-band weather radar data under strong convective storm precipitation. The proposed method achieves far smaller root mean square errors of the radar reflectivity factor than traditional attenuation correction methods in both the radial and the vertical directions. Overall, the proposed method has a better attenuation correction effect for C-band weather radar reflectivity data than traditional attenuation correction methods, which may enable the popularization and application of the proposed method elsewhere.

Although there are many state-of-the-art methods for hyperspectral classification, data deficiency is a problem that should be addressed before popularizing hyperspectral technology. To solve this problem, it is worth exploring methods based on small datasets. Inspired by the advanced deep learning classification methods and the autoencoder structure, we propose a structure named three-dimensional convolutional adversarial autoencoder that combines the two processes for semisupervised hyperspectral classification. Our experiments show its utility in data-deficient situations, and our study analyzes its advantages and disadvantages, and points out a probable direction toward optimization.

Calibration of the on-orbit gain changes of the narrow bandwidth reflective solar bands (RSB) of Terra and Aqua MODIS is usually based on the band center wavelength. The relative spectral response (RSR) of each band is assumed to be constant on orbit, and the time dependence of an overall gain factor is calculated. Any on-orbit changes to the RSR of the MODIS bands will introduce some error into the calibration and may also have an impact on the Earth scene radiance retrieval. We consider two different ways to track how the RSR of the MODIS RSB may change on orbit and the effects that these changes have on the calibration. First, we study the broadband degradation of the MODIS scan mirror and how it may change the effective out-of-band response of the RSB. Second, we examine in-band RSR measurements from the spectroradiometric calibration assembly carried on-board both MODIS instruments. We find that RSR changes have a relatively small effect on the radiance calibrated using the on-board solar diffuser, generally <0.5  %   for all RSB at any time in the missions, though the effect may be slightly larger for some scan angles. The impact of RSR changes on the Earth scene radiance retrieval is highly dependent on the spectral properties of the scene and could be significantly larger. The bands showing the largest impacts are Terra bands 1, 4, 8, 9, 17, and 19 and Aqua bands 8 and 9.

Paddy transplantation rate mapping can provide important information regarding water demand for its transplantation. It is a time-tagged activity that cannot be accurately mapped using a single date image. Further, many times, single sensor data may not have enough temporal resolution required to map it. Furthermore, due to cloud cover, a single optical sensor is not enough to provide required temporal scenes for mapping the transplanted paddy fields. Therefore, to have a finer temporal resolution, a multisensor approach was applied in our research work by integrating an optical sensor data with a synthetic aperture radar data to map transplanted paddy fields at a frequency of 5- to 10-day intervals. In our research, the transplanted paddy fields were mapped as a specific class of interest using suitable fuzzy-based classifiers. The specific fuzzy-based classifiers explored here were modified possibilistic c-mean (MPCM) and noise clustering (NC). For evaluating the results, mean membership difference (MMD) and entropy values were computed within transplanted paddy fields and for noninterest classes. Moreover, an error matrix criterion was also used to validate the results. The computed MMD and entropy values were found to be lower within transplanted paddy fields in comparison to the other classes of noninterest, indicating accurate classification. NC algorithm outperformed the MPCM algorithm while observing MMD and entropy values for assessment. However, using the fuzzy error matrix criterion, the performance of both algorithms was identical with an average overall accuracy and kappa coefficient of 90.88% and 0.82 for MPCM and 90.66% and 0.81 for NC, respectively.

By combining the data of different spatially distributed sensors, multistatic inverse synthetic aperture radar (ISAR) can provide more stable imaging results compared with monostatic ISAR. In most previous studies of multistatic ISAR imaging, the scatterers on the target are modeled as isotropic points and a conventional monostatic ISAR imaging method is directly applied after rearranging the data of multiple sensors, which will result in the degradation of image quality, especially when the measurements are limited. In practice, the complex amplitude of the scatterer is usually strongly angle dependent; therefore, the echo of different sensors cannot be assumed to be coherent. We consider the fluctuation of the radar cross section and propose a method based on compressed sensing (CS) for multistatic ISAR imaging. By utilizing the block orthogonal matching pursuit (BOMP) method to reconstruct the target image, the requirement of coherence between different sensors can be eliminated. Moreover, in order to apply the CS method, a two-step target motion estimation approach is also presented. A coarse motion estimation method is first applied by tracking the target trajectory with the distance-sum measurement. Then, associating the gradient-based optimization algorithm with BOMP, a parametric block-sparse reconstruction method is developed to jointly correct the residual position error and recover the target image with incoherent echo. Simulation results show the effectiveness of the proposed method.

Quantifying Geostationary Operational Environmental Satellite R-series (GOES-R) Advanced Baseline Imager (ABI) Channel (Ch) 2 low-light signal-to-noise ratio (SNR) is a core ABI postlaunch product test to validate prelaunch performance expectations of this parameter. For GOES-16, an SNR analysis has been performed to accomplish this goal using ABI Mode-3 MESO visible near-infrared (VNIR) channel L1b radiance product images taken every 30 s between 17:00 UTC and 17:15 UTC on May 23, 2017. Scene pixel data used in the analysis are first screened to ensure they meet the low-light criteria and to minimize scene temporal variability associated with meteorological evolution and geolocation uncertainty in highly inhomogeneous scenes. After such screening, image-to-image radiance time difference statistics are compiled and analyzed to estimate instrument SNR. Based on GOES-R ABI vendor prelaunch testing, minimum and mean expected ABI Ch 2 postlaunch SNR performance at 5% albedo was estimated to be 44.2 and 64.5, respectively. Meanwhile, the SNR computed in this analysis is found to be about 57  ±  8. Thus, low-light SNR for ABI Ch 2 is shown to meet minimum user expectations. The analysis for the other five GOES-16 ABI VNIR channels is also shown to contrast the SNR for these channels.

Providing microclimate conditions in greenhouse systems is very important in terms of growing desired products. In addition, remote sensing and data analysis processes need to be developed and expanded to perform fast and perfect monitoring and control in these systems. An application of remote sensing with infrared imaging in greenhouse systems has been developed. It is proposed to obtain the mean temperature inside the greenhouse via infrared images, the air relative humidity with psychrometric chart application, and the soil moisture values with the multiple regression analysis method. The control and monitoring of automatic irrigation, heating, humectation, ventilation, and lighting systems were carried out via the MATLAB–GUI interface. The desired conditions for the carnation plant were ensured in the greenhouse, and 24-h data were obtained. The digital and visual application findings showed that the designed system is more advantageous in terms of ease of use, monitoring, and simultaneous recording compared with the traditional sensor control systems.

To address the insufficiency of texture information-based classification features to classify samples, we proposed two methods for spatial information-enhanced hyperspectral imagery classification based on joint spatial-aware collaborative representation (JSaCR). First, we introduce a texture regularized-based joint spatial-aware collaborative representation (TRJSaCR) method, in which prior texture is regarded as a regularization term to constrain the coefficient of the objection function of JSaCR and the closed-form solution is obtained to reconstruct the test sample. Second is a spatial information-assisted discrimination rules (SIDR) method coupled with TRJSaCR (TRJSaCR-SIDR) for classification. More precisely, the label information of the test samples and their corresponding neighborhoods are specified by TRJSaCR-SIDR, and the final labels are determined by considering their neighborhood label distribution. Our work aims to broaden the knowledge of the utilization of spatial information in hyperspectral classification. Experimental results on two benchmark hyperspectral datasets, Indian Pines and Pavia University, indicate that the proposed algorithms are superior to other state-of-the-art classifiers.

Feature extraction (FE) is an important preprocessing step for classification of high-dimensional data. However, when the number of labeled training samples is small, less discriminant information induces some FE methods that use statistical moments, for example, linear discriminant analysis, even fail to work. We present a collaborative representation (CR)-based semisupervised FE method using attraction points (CRSUAP). By CR, CRSUAP defines a membership matrix that contains the correlation between multiple samples. Then, a nonmembership matrix is designed to enrich the discriminant information and further enhance the separability of classes. To avoid estimating statistical moments, an attraction point is selected from each class to calculate the projection matrix, avoiding matrix singularity problem. The experimental results on three real hyperspectral images demonstrate that CRSUAP has better performance than other related FE methods in a small labeled sample size situation.

Ship detection from synthetic aperture radar (SAR) images plays an important role in marine safety and ocean resource management. Unsupervised ship detection methods have a complex set of rules, while supervised methods, such as deep learning approaches, consume substantial time and manpower to make training samples. We demonstrate that ships in an SAR image can be detected by a weakly supervised convolutional neural network that combines new deep learning technology called class activation mapping with the conditional random field. Our model is trained using only SAR images with two global labels, namely, “ship” and “nonship,” and produces three types of output: ship location heatmap, ship bounding box, and pixel-level segmentation product. Experiments on Chinese Gaofen-3 fine strip SAR images validate the effectiveness of the proposed method. Compared with the state-of-the-art methods, our method achieves higher detection accuracy and more intelligent detection characteristics.

The precise and efficient location of residential areas using high spatial resolution remote sensing imagery is a popular research area in the field of Earth observation. Most of the existing approaches are supervised or semisupervised and use data training. Among the unsupervised approaches, corner density-based mapping using kernel density estimate has been widely employed to predict the presence of built-up areas. However, it is computationally time-consuming and the statistical threshold segmentation makes it difficult to obtain a stable and accurate output. To overcome this deficiency, a new two-stage object-oriented residential area extraction scheme was designed. First, a set of corners was extracted using the Gabor filter bank with structural tensor analysis to indicate candidate buildings. Then, instead of pixel units, our method takes superpixel-based image partitions as the primary calculation elements, and an object-oriented weighted sparse spatial voting technique was proposed to accelerate the generation of a residential area presence index. It was demonstrated that the superpixel-based voting strategy was not only efficient in accelerating the calculation process, but it also reduced the false negative rate in the final detection result. Second, a graph-cut method was employed to address the residential area segmentation by integrating a density map as a prior cue that preserves the boundary accuracy better than traditional statistical threshold methods. The effectiveness of the proposed method was evaluated using a series of experiments on the sets of high-resolution Google Earth, IKONOS, and GaoFen-2 (GF2) satellite imagery. The results showed that the proposed approach outperforms the existing algorithms in terms of computational speed and accuracy.

Metaheuristic algorithms have been widely used in determining optimum rational polynomial coefficients (RPCs). By eliminating a number of unnecessary RPCs, these algorithms increase the accuracy of geometric correction of high-resolution satellite images. To this end, these algorithms use ordinary least squares and a number of ground control points (GCPs) to estimate RPCs. Due to the cost of GCP collection, using limited GCPs has become an attractive topic in various research studies. A configuration for RPC estimation using metaheuristic algorithms, namely, discrete-binary configuration for rational function model (DBRFM), is presented to find the optimal number and combination of RPCs in the case of limited GCPs. Based on the fact that the maximum number of RPCs is twice the number of GCPs, the particle/chromosome in the proposed configuration is composed of two binary and discrete parts. This configuration not only is compatible with the nature of the metaheuristic algorithms but also significantly reduces the search space. The proposed method has been tested on various types of remotely sensed data sets. The results of the experiments indicate the superiority of the DBRFM in comparison with the traditional binary configuration of metaheuristic algorithms.

Since sea clutter samples are often contaminated by high-intensity outliers, target detection in nonhomogeneous sea clutter environments is challenging. An adaptively truncated clutter-statistics-based variability index constant false-alarm-rate (TSVI-CFAR) detector of Gaussian clutter in synthetic aperture radar (SAR) imagery is proposed. The proposed method is designed to improve the CFAR detection performance in a heterogeneous multitarget environment. TSVI-CFAR consists of three stages, i.e., clutter truncation, statistical parameter estimation, and CFAR detection. In the clutter truncation stage, the high-intensity outliers in the background window are eliminated through an adaptive threshold-based clutter truncation method. In the statistical parameter estimation stage, the parameters can be accurately estimated through the maximum-likelihood estimator. In the detection stage, the clutter background is categorized using the adaptively truncated clutter. Different types of CFAR detection methods are applied to the pixels under test of the categorized backgrounds. TSVI-CFAR has a higher detection rate and a low observed false alarm rate. The effectiveness of the proposed algorithm is demonstrated using Gaofen-3 SAR data and Envisat-ASAR data.

We propose exploratory research combining synthetic aperture radar (SAR) data, represented by Sentinel-1A, and multispectral data, represented by Landsat-8 operational land imager (OLI), to demonstrate the applicability and effectiveness of land cover classification based on a Beijing case study. The proposed method consists of two phases. In the fusion phase, we select three methods to evaluate the performance of integrated Sentinel-1A and Landsat-8 OLI images. In the classification phase, we choose four common methods to examine the classifying capability hidden within the fused images. Experimental results indicate that the Gram-Schmidt spectral sharpening is superior in terms of maintaining the geometric structure, spectral texture, and spatial information, demonstrating a better fusion effect than other methods. The support vector machine classification exhibits the best performance of the fused images, with an overall classification accuracy of 94.01% and a kappa coefficient of 0.91. The fused images provide better classification potential as they benefit from having more spatial information and spectral information distribution, and the mean value of overall classification accuracy and the kappa coefficient are on average 5.61% and 0.08 higher, respectively, than the original Landsat-8. Finally, we conclude that the integrated use of SAR and multispectral images significantly improves classification accuracies, thus making it effective for land cover information extraction.

The ultrahigh resolution of unmanned aerial vehicle (UAV) remote sensing images and tilting photography with multiple perspectives provide complete and detailed ground observation data for various engineering applications. However, noise and interference information make learning the typical features of ground objects difficult for current deep learning semantic segmentation networks. The hierarchical cognitive structure of human vision and the information transmission modes of retinal cone and rod cells were used to design a two-pathway anti-interference network for retinal perception mechanism simulation (RPMS). In the first pathway, the hierarchical cognition of cone cells was simulated by a one-to-one connected multiscale dilated convolution structure. In the second pathway, the hierarchical cognition of rod cells was simulated by a multiscale pyramid structure with many-to-one connections. With the one-to-one connection, the ability of RPMS to recognize detailed edges was strengthened. Furthermore, the many-to-one connection helped RPMS resist the disturbance from noise and interference. By combining the feature maps of the two paths, RPMS exhibited stronger noise resistance, better texture recognition, and better detail recognition compared with other semantic segmentation networks in the classification experiments. Thus this technique is suitable for UAV remote sensing image classification and has a broad application potential.

Entropy-based parametric autofocus is commonly applied to correct the residual azimuth phase error (APE) that arises from the residual motion errors (RME) in synthetic aperture radar. However, the image’s entropy trends to a convex function only when a great number of range bins are selected according to the law of large numbers, which means that the average residual APE is finally obtained via this method for the space-variant characteristics of RME. A two-stage parametric autofocus algorithm is carried out to derive the exact residual APE with several range bins selected for iteration. To this end, a new objective function in the first stage is designed to reduce the number of local solutions and protect targets from energy leaking. In the second stage, the filled function is imported to jump out of the current solution to a more optimal one until no other optimal solution exists. In the final part, both simulated and real data experiments validate the performance of the proposed method.

Fusion results of low-resolution multispectral (LRMS) images and high-resolution panchromatic (Pan) images, also called pan-sharpened images, are always difficult to evaluate due to the lack of high-resolution multispectral (HRMS) images and the complexity of the fusion process. By taking spectral information of LRMS images and spatial structural information of Pan images as references, we extract the saturation map and luminance value as spectral features, and construct the optimal contrast map and structure similarity map as spatial features to compute the four indices between the original LRMS, Pan images, and the fused result: saturation similarity, luminance consistency, contrast similarity, and structure similarity to describe distortions from different aspects. Then, we feed the four indices into an extreme learning machine to train a nonlinear pooling strategy, and finally a multifeature and learning-based model is constructed for fusion image quality assessment. Comparisons with state-of-the-art image quality assessment metrics show that the proposed metric gains a much higher consistency with subjective opinions while needing no reference HRMS images.

Airborne hyperspectral images used for remote sensing are distorted by various factors, such as atmospheric effects, transmission noise, instrumentation noise, and motion blurring. Proper assessment of image quality is extremely important in the identification and characterization of distortion, evaluation of compression performance, and so on. We present an ensemble feature-based full-referenced approach to quantify the quality of remotely sensed hyperspectral images. Our ensemble features quantify the objective quality of the image inconsistency with the visual measure and identify the inherent distortions. The proposed approach identifies the distinct spatial structural image features from the images corresponding to each spectral band and obtains the hyperspectral cube quality by computing the mean. The measure also identifies the highly distorted spectral bands, which must be restored or eliminated before processing. We evaluate objective image quality in several real hyperspectral images and conclude that our proposed approach evaluates the image quality more efficiently compared to the existing approaches.

Spectral variability is an inevitable problem in spectral unmixing. The linear mixing model (LMM) is often used due to its simplicity and mathematical tractability. Unfortunately, the linear mixing assumption is not always true in many real scenarios. To address this issue, we adopt a variant of the LMM to take care of spectral variability, called scaled and perturbed LMM, which can be used to constrain modeling reflectance scaling caused by topography or illumination, and simulating irregular spectral variabilities. To facilitate effective optimizations of the variables, a few regularizations are employed to regularize the introduced constraints and an alternating direction method of multipliers algorithm is further used to optimize all the variables of this model. Experimental results obtained from a synthetic dataset and two real datasets demonstrate that the proposed approach outperforms other algorithms in unmixing hyperspectral images with spectral variabilities.

Deep learning has been widely used in hyperspectral image (HSI) classification. However, a deep learning model is a data-driven machine learning method, and collecting labeled data is quite time-consuming for an HSI classification task, which means that a deep learning model needs a lot of labeled data and cannot deal with the small sample problem. We explore the small sample classification problem of HSI with graph convolutional network (GCN). First, HSI with a small number of labeled samples are treated as a graph. Then, the GCN (an efficient variant of convolutional neural networks) operates directly on the graph constructed from the HSI. GCN utilizes the adjacency nodes in graph to approximate the convolution. In other words, graph convolution can use both labeled and unlabeled nodes. Therefore, our method is a semisupervised method. Three HSI are used to assess the performance of the proposed method. The experimental results show that the proposed method outperforms the traditional semisupervised methods and advanced deep learning methods.

When the micromotion (MM) target has a large MM amplitude, the bandwidth of the synthetic aperture radar (SAR) azimuth echo will be greater than the pulse repetition frequency (PRF) of SAR. According to Nyquist sampling theorem, spectrum aliasing will occur at this time, and conventional MM target detection and estimation algorithms may lose effect. To solve the problem of large MM target detection with echo bandwidth greater than the PRF, an inverse Radon transform (IRT)-based large MM target detection algorithm is proposed. The study found that the time-frequency (TF) spectrum of large MM target azimuth echo is usually a folded sinusoidal curve. Then we spliced enough identical TF spectrums to restore an intact sinusoidal curve. The intact sinusoidal curve can be mapped into a peak in the parameter space using IRT. Hence, the algorithm can detect large MM target and estimate its parameters accurately. Data processing results prove the effectiveness of the algorithm. At the same time, the performance analysis proves that the operation speed and antinoise performance of the algorithm are better than that of the Hough transform algorithm.

Speckle noise can reduce the image quality of synthetic aperture radar (SAR) and make interpretation more difficult. Existing SAR image despeckling convolutional neural networks require quantities of noisy–clean image pairs. However, obtaining clean SAR images is very difficult. Because continuous convolution and pooling operations result in losing many informational details while extracting the deep features of the SAR image, the quality of recovered clean images becomes worse. Therefore, we propose a despeckling network called multiscale dilated residual U-Net (MDRU-Net). The MDRU-Net can be trained directly using noisy–noisy image pairs without clean data. To protect more SAR image details, we design five multiscale dilated convolution modules that extract and fuse multiscale features. Considering that the deep and shallow features are very distinct in fusion, we design different dilation residual skip connections, which make features at the same level have the same convolution operations. Afterward, we present an effective L__hybrid loss function that can effectively improve the network stability and suppress artifacts in the predicted clean SAR image. Compared with the state-of-the-art despeckling algorithms, the proposed MDRU-Net achieves a significant improvement in several key metrics.

The matting technique can accurately detect the differences between the foreground and the background of images, but most of the existing matting algorithms need foreground and background labels in advance, which leads to obvious limitations in the processing of mass images. With matting technology introduced into cloud detection for multisource satellite images, an automatic and accurate cloud detection algorithm is proposed. First, the S-curve is used for the original image to enhance the contrast between the cloud and the background, then the gray visible image is binarized by the ensemble threshold methods and a preliminary binarized map is obtained by the voting strategy. On this basis, the connected regions are calculated and the centers of the connected regions are taken as the seed point, then the trimap based on the connected regions is automatically generated. Finally, the improved Laplacian matrix and the conjugate gradient algorithm are used to solve the alpha value to obtain accurate results of cloud detection. The proposed method is compared with learning-based and robust matting-based cloud detection algorithms. Objective and subjective experiments based on images from the Fengyun-2G meteorological satellite, Landsat-8, and Sentinel-2 show that this proposed method can detect multiple types of clouds better and exhibits higher accuracy.

For local intensity-based image registration methods, a template of a predefined size is adopted. Some measure of similarity between the images is used to determine when the optimal alignment has occurred. Most intensity-based similarity utilizes the statistical and spatial information of an image. However, there are significant nonlinear radiometric differences between infrared and visible data. Images of natural scenes usually do not have enough spatial features within the template. Hence, many similarities will be useless when dealing with infrared and visible data. Minor eigenvalues (ME) image registration similarity is presented with the exploitation the spectral properties of remote sensing images. ME similarity is based on the linear spectral mixture model and detects control points through searching the minimum of ME of the covariance matrix. Experiments on Landsat-7 satellite Enhanced Thematic Mapper Plus data are performed to verify and evaluate the effectiveness. Transformation performance curves, correct match ratio (CMR), and registration accuracy are also discussed. According to the data, root-mean-square error of phase correlations is 0.0722 pixels and the CMR of ME similarity is nearly 100%. The results on the basis of TM1, TM2, TM3, and TM4 band images indicate that the proposed similarity holds promise for infrared and color image registration in natural scenes, with advantages over previous normal mutual information and gradient mutual information similarities.

With the advances in remote sensing, wild animals sprawling over a vast area can be easily and quickly captured using low-cost unmanned aerial vehicle imagery. We propose an aerial animal detection and counting network (DetCountNet) framework called FSSCaps-DetCountNet, using fuzzy soft sets (FSS) and capsule network (CapsNet). Similarity measures based on FSS have been used to discriminate the target animals from both nontargets and the background. Of particular interest to aerial images, CapsNet requires very few training data and is robust to rotation and affine transformation. With superpixel segmentation and attention maps, FSSCaps-DetCountNet works well on challenging image conditions, such as dense background with sparse animals and overlapping/cluttered animals. The model is trained and tested on benchmark aerial animal datasets, namely, the aerial elephants and the livestock datasets with an accuracy index of 99.84% and 99.86%, respectively. Also, the overall omission and commission errors are 0.02% and 0.03%, respectively. The experimental results and comparative study with other state-of-the-art conventional models demonstrate the effectiveness and robustness of FSSCaps-DetCountNet for real-time animal detection and counting from aerial images.

To improve the performance of exiting sparse recovery-based space-time adaptive processing (STAP) methods in nonstationary and heterogeneous environments, a STAP method, called the joint sparse representation (JSR)-based direct data domain STAP method using MIMO subarraying radar, is proposed. First, by exploiting the property of waveforms orthogonal diversity of MIMO subarraying radar, the received single snapshot signal of the range cell under test is expressed as multiple snapshot signals. Second, the JSR model of clutter plus target spatiotemporal spectrum estimation is established and the joint sparsity is proved. The spatiotemporal spectrum of clutter and target is then estimated by the typical JSR algorithm. At last, the clutter covariance clutter matrix and the corresponding weight vector of the STAP processor are calculated based on the clutter space-time distribution extracted by the rough priori knowledge of the target. By utilizing the data of the range cell under test only, the proposed method could achieve accurate spatiotemporal spectrum estimation and great performance of clutter suppression. Experimental simulations demonstrate the validity of the proposed method.

Pan-sharpening is an important image preprocessing technique for remote sensing that aims to enhance spatial resolution of multispectral (MS) images under the guidance of panchromatic (PAN) image while preserving spectral properties. The existing pan-sharpening methods usually adopt the globally consistent detail-injection models, neglecting the detail differences between spectral channels, which leads to imprecise spatial details and distorted spectral properties of pan-sharpening results. We propose a sparse representation-based detail-injection model for pan-sharpening that utilizes the structure similarity and detail differences between PAN and low-resolution multispectral (LRM) images at each channel, to improve the performance of pan-sharpening. Specifically, to better express the inherent detail properties of the MS image, the overcomplete dictionary of each channel is constructed from synthesized high-resolution multispectral (HRM) images. Moreover, the most proposed methods require that the spectral responses of the PAN image and the MS image cover the same wavelength range; nevertheless, most sensors cannot match this condition. To address this problem, we propose constructing coupled low-resolution and high-resolution dictionaries from LRM and synthesized HRM images so that the structure similarities can be used for detail injection. The qualitative and quantitative experimental results on various data sets demonstrate the superiority of our proposed method over the state-of-the-art methods.

The aim of this study is to investigate the potential of the Sentinel-2 satellite for land use and land cover (LULC) mapping. The commonly known supervised classification algorithms, support vector machines (SVMs), random forest (RF), and maximum likelihood (ML) classification are adopted for investigation along with a proposed hierarchical classification model based on a coordination of information on the environment land cover system. The main classes for land cover and mapping in the proposed hierarchical classification are selected as water, vegetation, built-up, and bare land in the first level, which is followed by inland water, marine water, forest/meadow, vegetated agricultural land, barren land, and nonvegetated agricultural land in the second level. The study areas for the experiments are selected as the two biggest cities of Turkey, namely Ankara and Izmir, providing a sufficient number of classes for comparison purposes. During the utilized hierarchical methodology, water and vegetation are first extracted using the normalized difference water and vegetation indices. This is followed by the selection of training pixels from the remaining classes to perform and compare different supervised learning algorithms for the first- and second-level classification in terms of accuracy rates. The experimental results first reveal that while the SVMs have close accuracy performances to those with RF, they are significantly superior to the ML classification, with an average of 8% accuracy rates for LULC mapping. Second, the hierarchical classification also gives higher performances with respect to the nonhierarchical classification, with the provided gains between 4% and 10% for class-based accuracies. The overall accuracy rates of the proposed hierarchical methodology are 85% and 84% for the first-level classes and 83% and 72% for the second-level classes, respectively, for Izmir and Ankara.

Long-term coherent integration is an effective method for target detection. However, the inevitable motion compensation errors and complex background clutter seriously limit the detection performance of a cell-averaging constant false alarm rate (CFAR) detector. A generalized CFAR detector on the Riemannian manifold of Hermitian positive-definite matrix is proposed. Based on the α log-determinant divergence (α-LDD) and symmetrized α-LDD, the mean matrix of the reference cells is first estimated. Then the dissimilarity between the covariance matrix of the cell under test and the mean matrix is measured to determine the presence of a target. Numerical experiments show that the proposed detectors have better detection performance than other representative methods when detecting a target with motion compensation errors in complex background.

When a satellite sensor with a large field of view and wide swath is calibrated, it is not easy to obtain the image when the calibration site is located precisely at the nadir position. If the location of a calibration site is at an off-nadir position in the image, calibration errors will be caused by the inconsistent observation angle between the sensor view and the ground measurement view. The bidirectional reflectance distribution function (BRDF) model plays an important role in solving this problem. In this study, a BRDF measurement system based on an unmanned aerial vehicle (UAV) is developed. This system has the capability of measuring angular data with observation azimuth angle ranging from 0 deg to 360 deg with an angle interval of 30 deg, and observation zenith angle ranging from 0 deg to 50 deg with an angle interval of 10 deg. The directional data of the Dunhuang calibration site were measured using the UAV BRDF measuring system at different solar zenith and azimuth angles, and the spatiotemporal distribution characteristic of forward- and backward-scattering of Dunhuang calibration site was analyzed. A Ross–Li BRDF model, built using measurement data, is used to calculate the directional surface reflectance under any observation geometry of solar and satellite. These calculations are applied to correct the calibration data of the CBERS-04 WFI sensor. Results show that the BRDF model significantly improves the calibration accuracy, especially in the case of large observation angles.

Waveform decomposition using the Levenberg–Marquardt algorithm is a powerful tool for detecting surface return, bottom return, and volume backscatter return in a superposed green waveform of airborne LiDAR bathymetry (ALB). However, traditional decomposition methods do not handle bound constraints and are easily trapped in local optimum. Thus, the decomposed components may be inconsistent with the measurement principle of ALB. This study proposes an improved waveform decomposition method by setting reasonable lower and upper bounds of waveform parameters to guarantee the fidelity of the decomposed components. First, a comprehensive mathematical model of a green waveform is proposed by considering the early return. Second, the lower and upper bounds of the waveform parameters are given on the basis of the measurement principle of ALB. Finally, improved waveform decomposition is achieved using the comprehensive model and a constrained nonlinear optimization. The proposed method is applied to a practical ALB measurement using Optech coastal zone mapping and imaging LiDAR. Compared with traditional decomposition methods, the improved waveform decomposition not only ensures good fitness but also guarantees the fidelity of the decomposed components.

The utility of the ground-based soil observation laser absorption spectrometer (SOLAS) was demonstrated through spectral reflectance measurements of five soil types during laboratory- and field-based collection. The SOLAS telescope-assisted measurements were compared with proximal measurements. The spectra that were acquired at an intermediate range of 40 meters compared well with the spectra that were acquired proximally. Specimen type, range-dependent spatial resolution, and environmental conditions are discussed. The signal-to-noise ratio (SNR) was assessed and is presented as a function of wavelength for the spectral range of the receiver for each measurement condition. The proximal measurements performed outdoors under solar illumination had the greatest SNR, while the remote measurements performed indoors under artificial illumination had the lowest SNR. For the outdoor measurements, loss of signal was observed around the 1400- and 1900-nm bands due to long-path atmospheric water vapor absorption. The discussed future improvements to the SOLAS remote sensor will enable measurements of reflectance over longer ranges. Envisioned applications include remote characterization of surface materials for large construction projects (e.g., surface mines and tailings) for geohazard investigations, or for ground truthing of current and future multispectral and hyperspectral satellite data.

In the agricultural field, optical remote sensing technology plays an important role in crop monitoring or production estimation. However, the widespread distribution of clouds and rain limits the application of optical remote sensing. Synthetic aperture radar (SAR) has been widely used for studies of oceans, atmosphere, land, and space exploration, as well as by the military due to its all-weather nature, penetration to surface and cloud layers, and diversity of information carriers. However, it is difficult to classify ground objects with high accuracy based on SAR data. Considering the features of these two datasets, we proposed a framework to improve crop classifications in cloudy and rainy areas based on the optical-SAR response mechanism. Specifically, this method is designed to train a parametric analytic model in the area using both kinds of datasets and applied in the area with only SAR data to obtain the optical time-series features. Then crops from the second area were classified by the long-short-term memory network. As an example, the parametric analytic model in Lixian County was studied and was applied to Xifeng County to classify the crops with the OA of 61%, which had proved the robustness of the method.