This paper presents a frequency modulated continuous wave (FMCW) technology-based airborne synthetic aperture radar (SAR) system. This radar system operates on an eight-channel simultaneous dual band (X- and L-bands) in fully polarimetric mode. Compared with the traditional multiband airborne SAR, this radar system features a small size and low power consumption based on the FMCW technology. The antenna subsystem can be easily mounted behind the cargo door of a small aircraft platform without any modifications to the aircraft structure. A Cessna 208 is employed to mount this radar system in a left side looking installation. A single pass flight test is carried out in China, and preliminary imaging results are acquired and classified. The merged X–L band image provides the highest classification accuracy compared with the single X- and L-bands and optical images. The experimental results indicate that the proposed FMCW SAR is a powerful tool for terrain classification, especially for rural-urban region classification.

The Moderate Resolution Imaging Spectroradiometer (MODIS) instruments have successfully operated for more than 18 and 16 years, respectively, on-board the NASA’s Earth Observing System Terra and Aqua spacecraft. Both Terra and Aqua MODIS have significantly contributed to the advance of global Earth remote sensing applications with a broad range of science products that have been continuously produced since the beginning of each mission and freely distributed to users worldwide. MODIS collects data in 20 reflective solar bands (RSB) and 16 thermal emissive bands (TEB), covering wavelengths from 0.41 to 14.4 μm. Its level 1B (L1B) data products, which provide the input for the MODIS high-level science products, include the top of the atmosphere reflectance factors for the RSB, radiances for both the RSB and TEB, and associated uncertainty indices (UI) at a pixel-by-pixel level. This paper provides a brief review of MODIS L1B calibration algorithms, including a number of improvements made in recent years. It presents an update of sensor calibration uncertainty assessments with a focus on several new contributors resulting from on-orbit changes in sensor characteristics, approaches developed to address these changes, and the impact due to on-orbit changes on the L1B data quality. Also discussed are remaining challenges and potential improvements to be made to continuously maintain sensor calibration and data quality, particularly those related to the quality of MODIS L1B uncertainty.

In response to the increasing threat of terrorism, standoff security screening of person-borne concealed weapons or contraband in public areas, such as airports and railway stations, is an urgent requirement; however, current solutions suffer from the limited field of view and the problem of security screening with interfering with the passenger flow. This paper proposes an imaging method that exploits the angular diversity of distributed multistatic terahertz (THz) radar network in combination with inverse synthetic aperture radar (ISAR) technique to realize high-resolution penetration imaging at a standoff range. The radar network is based on multistatic ISAR systems, and the multistatic ISAR system’s equivalent monostatic model is derived. According to the equivalent monostatic model, we develop the image formation algorithm and give the analytical expression of the cross-range resolution. Compared with conventional monostatic ISAR configuration, the analytical expression shows that the multistatic ISAR configuration can enhance the image’s cross-range resolution and the maximum enhanced factor is approximately equal to the number of radars in the multistatic system. Numerical simulations and experimental results agree well with the theoretical derivation. The proposed imaging approach offers a low-cost and flexible alternative to the present systems not only for security screening but also for noncooperative target imaging.

Time delay and integration (TDI) detectors are preferred image sensors for high-resolution spaceborne imaging systems for cartographic applications as they provide higher sensitivity by integrating charges over multiple TDI stages in push broom mode. Large swath requirements are met using multiple TDI detectors in optical butting configuration in the focal plane assembly (FPA) to form a single image line on the ground. These detectors need to be aligned with subpixel accuracy to avoid motion smear and errors in the image mosaicking process. Alignment of the multiple TDI detectors in the FPA is a challenging task and is accomplished using custom methods and setups. Passive alignment techniques rely on very high-resolution microscopes for achieving desired accuracies. Active method requires TDI detectors to be operated in staring mode, which is architecturally and functionally not supported by TDI detectors. We developed a scheme for operating the TDI detectors in staring mode without any smear effects to aid in the active alignment process. Based on this, we propose an active alignment approach. The proposed approach is highly advantageous as it can be used throughout the imaging system development phase to demonstrate stability of the alignment and also helps in cross validation with the passive alignment. The developed method was test validated on qualification model FPA, and alignment accuracy of the order of 0.4  μm in tilt and 1.2  μm in line shift has been demonstrated, against the requirements of ±1.5  μm. This method has been extensively used in Cartosat-2S series imaging systems and will be very useful for upcoming high-resolution missions of the Indian Space Research Organisation.

The present study is designed to explore the potential of bistatic scattering coefficients (σ  °  ) and machine learning algorithms for the estimation of rice crop variables using ground-based multiangular, multitemporal, and dual-polarized bistatic scatterometer data. The bistatic scatterometer measurements are carried out at eight different growth stages of the rice crop in the angular range of incidence angle 20 deg to 70 deg for HH- and VV-polarization at 10-GHz frequency in the specular direction with an azimuthal angle (φ  =  0). Several field measurements are taken for the measurement of rice crop variables, such as vegetation water content, leaf area index, and plant height at its various growth stages. Machine learning algorithms—such as fuzzy inference system (FIS), support vector machine for regression (SVR), and generalized linear model (GLM)—are used to estimate the rice crop variables using bistatic scatterometer data. The linear regression analysis is carried out for the evaluation of the multiangular, multitemporal, and dual-polarized datasets for the selection of optimum incidence angle and polarization for accurate estimation of rice crop variables. The highest value of the coefficient of determination (R2) is found at 30-deg incidence angle for VV-polarization. The sensitivity of copolarized ratio of σ  °   with the rice crop variable is also evaluated using linear regression analysis for the estimation of rice crop variables. The highest value of R2 is found to be at 35-deg incidence angle between the copolarized ratio of σ  °   and rice crop variables. The performance of SVR model is found superior in comparison to the FIS and GLM at VV-polarization and the copolarized ratio of σ  °   for the estimation of rice crop variables. However, the copolarized ratio of σ  °   is found superior to VV-polarized bistatic scatterometer data for the estimation of rice crop variables.

Scatterometers are microwave radars deployed to estimate global wind speeds and directions over the sea surface by measuring normalized radar cross-section (σo) of the sea surface. Measured σo is determined by the sea surface roughness that is modulated mostly by wind, thus establishing a relationship between σo and wind conditions. Wind vectors are retrieved by inverting an empirical geophysical model function that relates measured σo to a wind vector (speed and direction) and radar measurement configuration (polarization, incidence, and azimuth angle). Careful scatterometer calibration is required for accurate wind retrieval. NASA’s RapidScat scatterometer operated on board the International Space Station between September 2014 and August 2016. It was a twin instrument of the legacy QuikScat scatterometer. This paper proposes a method to cross-calibrate σo measurements between the RapidScat and QuikScat, taking advantage of their overlapping orbits. The σo biases between the two instruments have been calculated from observations collected during a 19-month period between January 2015 and August 2016. Deviations from the common reference model are computed for both datasets, as a function of wind speed, relative wind direction, and incidence angles averaged over 1000 pairs of collocated QuikScat/RapidScat revolutions.

The reflective solar bands (RSB) of the Visible and Infrared Imaging Radiometer Suite (VIIRS) on Suomi National Polar-orbiting Partnership satellite are calibrated using solar diffuser and lunar observations. To evaluate the performance of the VIIRS RSB calibration, the deep convective clouds (DCC) are utilized as invariant targets to assess the stability of NASA’s VIIRS sensor data record (SDR) reflectance product for 10 moderate resolution bands (M-bands, M1 to M5 and M7 to M11) and three imagery resolution bands (I-bands, I1 to I3). All the frames along with each scan are divided into six aggregation zones for the scan angle. An empirical bidirectional reflectance distribution function (BRDF) correction is performed to SDR reflectances over DCC to reduce the scattering anisotropy in view and solar angles. The BRDF-corrected reflectances are used to assess the stability and to explore the residual RVS effects for selected bands. Results of this study show that after more than 6 years of on-orbit operation, the use of the prelaunch-based RVS still meets the radiometric requirement for the VIIRS RSB. Examination of the small but noticeable RVS residuals indicates they occur in a few shortest wavelength bands based on a time-dependent quadratic fit of the reflectance trends obtained at all scan angle zones. As the mission continues, continuing monitoring of on-orbit RVS stability and its effects on VIIRS SDR reflectance products are necessary with the DCC technique in the future VIIRS calibrations to ensure the quality of the RSB SDR products.

Modern military targets as aircraft are able to perform high maneuvers due to their complex design. This ability of maneuvers includes sudden changes in acceleration and high-G turns that are not achievable from traditional military targets. Moreover, recent military targets generally have a low radar cross-section or low signal-to-noise ratio (SNR) profile, which makes the detection and tracking of those maneuvering targets a complicated dynamic state estimation problem. In this case, the track-before-detect filter (TBDF) that uses unthresholded measurements is considered as an effective method for tracking and detecting a single maneuvering target under low SNR conditions. Nevertheless, the performance of the algorithm will be affected by severe loss because of the mismatching of target model during maneuver. To resolve the target maneuvers, we propose an application of particle filtering, which depends on TBD. We employ the constant acceleration model and coordinate turn model. Our simulation results show that the detection and tracking of maneuvering target performance of TBDF has been improved using the proposed algorithm.

Deep learning has been applied to many fields due to its capability for unsupervised feature learning. However, it is unsuitable for remote sensing image classification of deep convolutional networks due to the regular and fixed shape of receptive fields in current methods, which cannot freely adapt for ground objects with various shapes and scales. To tackle this problem, we propose object-based deep convolutional autoencoders (ODCAEs) to encode high-resolution remote sensing image features automatically. The receptive fields of deep convolutional autoencoders were designed based on a fractal net evolution approach to adapt for various ground objects in images, retaining category purity while providing adequate information. We collected 109,333 samples of seven classes from high-resolution satellite images over various locations to train the ODCAE, which encoded high level features automatically, coupled with a support vector machine for classification. We assessed the classification results using two WorldView-II image locations. The proposed ODCAE approach achieved higher accuracy than three manual design feature systems. Thus, the proposed ODCAE approach is useful and efficient for feature learning problems for high-resolution remote sensing images.

Due to its excellent performance in terms of fast implementation, strong generalization capability, and straightforward solution, extreme learning machine (ELM) has attracted increasing attention in pattern recognition such as face recognition and hyperspectral image (HSI) classification. However, the performance of ELM for HSI classification remains a challenging problem especially in effective extraction of the featured information from the massive volume of data. To this end, we propose a method to combine convolutional neural network (CNN) with ELM (CNN–ELM) for HSI classification. As CNN has been successfully applied for feature extraction in different applications, the combined CNN–ELM approach aims to take advantages of these two techniques for improved classification of HSI. By preserving the spatial features while reconstructing the spectral features of HSI, the proposed CNN–ELM method can significantly improve the accuracy of HSI classification without increasing the computational complexity. Comprehensive experiments using three publicly available HSI datasets, Pavia University, Pavia center, and Salinas, have fully validated the improved performance of the proposed method when benchmarking with several state-of-the-art approaches.

GaoFen-3 (GF-3) is the first fully polarimetric C-band synthetic aperture radar (SAR) satellite in China. Interferometric observation is an important capability of GF-3. However, the long temporal baseline is a great challenge for repeat-pass interferometric synthetic aperture radar (InSAR) processing, especially over mountainous areas. Phase noise and complex topography make phase unwrapping difficult. This paper proposes a step-wise approach for high-resolution InSAR data processing focusing on facilitating phase unwrapping. First, the coarse GF-3 digital elevation model (DEM) is reconstructed by InSAR processing with multilooking of large looks, which is used to suppress phase noise. Second, the coarse GF-3 DEM is used to remove the terrain phase from the InSAR phase with multilooking of small looks, which can reduce phase gradient and improve the accuracy of phase unwrapping. After several iterations, the GF-3 DEM is finally inversed with multilooking of 2 looks, which is of high resolution and high accuracy. The performance of GF-3 C-band InSAR data on the extraction of DEM is examined over a deciduous forested area with the proposed method. The root-mean-square error of the extracted GF-3 DEM is 3.78 m in comparison with the LiDAR DEM.

An intrapulse polarimetric inverse synthetic aperture radar imaging method is proposed, under low signal-to-noise ratio (SNR), based on range profile fusion and improved motion compensation. Initially, the time delay and target movement between different polarized channels are compensated to enhance the echoes as full polarimetric data. Further, the high-resolution range profiles of different channels are fused to improve the SNR, using a polarimetric whitening filter, followed by the extraction of range bins containing the target region. Finally, improved motion compensation by joint range alignment and the respective phase adjustment of different channels are carried out to improve the compensation precision. Simulated and real measured data demonstrate the effectiveness of the proposed algorithm.

Nonrigid point set registration is a key technology in the field of remote sensing image registration, which is widely used in military and civil fields such as natural disaster damage assessment, agricultural and urban land-use planning, environmental quality monitoring, and ground target identification. We present a multifeature energy optimization framework and parameter adjustment-based nonrigid point set registration that has three contributions: (1) an energy optimization framework is designed to freely combine multiple features for estimating correspondences between two point sets, (2) the thin-plate spline and Gauss radial basis function transformation models can be optionally implemented for solving two-dimensional, three-dimensional, or higher-dimensional registration problems, and (3) the free parameters can be adjusted by the proposed three parameter adjustment approaches to yield higher registration accuracy in varied registration patterns. We test the performances of our method in contour point sets, sequence images, remote sensing images, medical images, and real images and compare with 10 state-of-the-art methods, where our method shows favorable performances in most scenarios.

Synthetic aperture radar (SAR) imagery requires image reproduction through the successive signal processing of received data before browsing images and extracting information. The received signal data records of ALOS-2/PALSAR-2 are stored in the onboard mission data storage and transmitted to the ground. To compensate for storage usage and transmission data capacity through the mission data communication network, the operation duty of PALSAR-2 is limited. This balance strongly relies on network availability. The observation operations of current spaceborne SAR systems are rigorously planned by simulating the mission data balance, given conflicting user demands. This problem should be solved so that we need not compromise the operations and potential of next-generation spaceborne SAR systems. One of the solutions is to compress the SAR data through onboard image reproduction and information extraction from the reproduced images. This is also beneficial for fast delivery of information products and event-driven observations by constellation. The Emergence Studio (Sōhatsu kōbō in Japanese) of the Japan Aerospace Exploration Agency has been developing evaluation models of the image processing system with the field-programmable gate array for onboard SAR image reproduction. The first model called the “Fast L1 Processor (FLIP)” can reproduce a 10-m resolution, single-look complex image (level 1.1) from ALOS/PALSAR raw signal data (level 1.0). FLIP processing speed at 200 MHz results in processing about five times faster than CPU-based computing at 3.7 GHz, corresponding to 50% real-time processing capability for ALOS/PALSAR. The FLIP’s image is almost identical to the image reproduced by 32-bit simulation in MATLAB.

The utilization of the complementary information, the color feature, and the intensity-texture feature, in hue, saturation, and intensity space is limited in the change detection of multispectral images. A kernel fusing the complementary information of the color feature and the intensity-texture features is proposed. To construct the fusion kernel, color kernel and intensity-texture kernel are constructed based on Gaussian radial basis function kernel. Then, the kernel cross-modal factor analysis method is used to provide transformations of the kernels and the fusion kernel is obtained by multiple kernel fusion, thus achieving the capture of the complementary information. In this way, the obtained fusion kernel highlights the changed areas and its boundary using the color features, and meanwhile provides texture-intensity change information. Experiments on real multispectral images demonstrate the effectiveness of the fusion kernel, and illustrate that it can get smoother homogeneous areas, obtain more accurate edge locations, and provide strong noise immunity for multispectral image change detection.

Hoz-e-Sultan playa is one of the lowest points in the center of Iran, and its size and shape rely on the amount of inlet and outlet discharged water in different seasons of the year. The temperature fluctuations are the significant variations in the surface of the playa that increases the maximum salinity and saline layer level. Therefore, the present study aimed to investigate the spatial-temporal changes and salinity rate obtained from 11 sedimentary cores in the study area by principal component analysis, Normalized Difference Salinity Index (NDSI), and Land Surface Temperature (LST) method. The results indicated that the level of playa decreased about 33.598  km² during 26 years. The spatial changes occurred in the eastern margin in the south of playa. Finally, a significant relationship was observed between the SI and the real salinity, and the NDSI and LST also increase the salinity rate related to temperature decrease.

Ship classification in synthetic aperture radar (SAR) images is essential in remote sensing but still full of challenges in the deep learning era. The unbalanced dataset and lack of models are two limitations. Upsampling with data augmentation and ratio batching are proposed to solve the first problem. Upsampling with data augmentation is upsampling by cropping and flipping. It can improve the diversity of the dataset. Ratio batching is realized by choosing the same amount of ships per class in each minibatch. It can make the model converge faster and better. To solve the second problem, a new loss function and convolutional neural network model are proposed. The new loss function can maximize the intraclass compactness and interclass separation simultaneously. Dense residual network has two submodules. One is the identity mapping through elementwise summation to reuse old features. The other is dense connection through concatenation to exploit new features. The designed architecture is suitable for the task of SAR ship classification. We use the confusion matrix and accuracy averaged on classes to measure the performance. From the experiments, we can find that the proposed ideas have excellent performance especially on the rare classes.

Inverse synthetic aperture radar (ISAR) is a powerful radar-processing technique that uses target’s motion to generate images on the range-Doppler plane. In the defense industry, ISAR imaging of moving targets is an important tool for automatic target recognition. We focus on the problem of ISAR imaging at low signal-to-noise ratio (SNR). The nonsubsampled directional filter bank (NSDFB) is a very useful tool in analyzing the directional information in two-dimensional signals. This paper presents an ISAR Imaging algorithm using NSDFB coefficients modeling. Bayesian maximum a posteriori is used where the heavy-tailed Lévy model is assumed for estimating an ISAR image at low SNR. We applied NSDFB transform to the ISAR image and developed a simulation procedure to describe the characteristics of the algorithm. Both simulated and real ISAR data have been tested. The proposed algorithm maintains a balance among noise suppression, feature preservation, and computational time. Finally, the experiments show that the proposed method outperforms others in terms of visual evaluation and image assessment parameters.

The speckle reduction problem for polarimetric synthetic aperture radar images is a complex issue due to difficulties on preserving polarimetric information. Several filters have been developed for this purpose such as Novak filter, enhanced Lee filter, and wavelet filters. We present an improved version of the Li filter by adding two false alarm detectors to enhance target appearance. These detectors are applied according to the homogeneity nature of the area to filter. The filtering process is performed by carrying a filtering step for each part of the SLC SAR image, intensity, and phase, independently from each other and using a contour detection technique to maintain the linear structures in the images. A comparative study with enhanced Lee filter is done and allows to conclude that introducing constant false alarm rate detectors gives the expected results: better targets (or reflectors) preservation, smoothing homogeneous regions, and details preservation.

A matrix-formed iteratively reweighted complex approximate message passing (MIRCAMP) method is proposed for inverse synthetic aperture radar (ISAR) imaging with two-dimensional (2-D) cluster sparse structure. The main idea of the proposed method is extending the complex approximate message passing into matrix-form and utilizing the iteratively reweighted way to exploit the 2-D cluster sparse structure in the ISAR image. The improved reconstruction performance of the MIRCAMP contains two parts: one is enhancing 2-D cluster signal reconstruction performance through a coupling relation between the sparsity patterns of neighboring coefficients. The other is improving recovery efficiency through 2-D matrix-formed iteration to avoid the large memory and computational requirements. Experimental results show the superior reconstruction performance of the proposed MIRCAMP.

This paper establishes the fuzzy autoencoder (FAE) to detect multiple changes between two one-dimensional multitemporal images. Different from the traditional approaches based on the pixel intensity, FAE includes a multilayer structure through self-reconstruction to extract the feature from an image. Due to the existence of noise in the images, the raw data tend to be corrupted and fail to detect the real changes. Therefore, the fuzzy number is introduced to the autoencoder to establish the FAE which is able to suppress the noise and learn robust features. In this way, the information in the fuzzy domain is introduced into the input, and in practice the fuzzy domain is discretized to facilitate the calculation. In addition, the weighted Frobenius norm is used to establish the loss function which can be minimized to achieve the optimal parameters. The framework is highlighted by the newly designed FAE. As the fuzzy number is introduced into the autoencoder, more information concerning the fuzzy domain is taken into consideration and thus the impact brought by the noise is relieved to a large extent. Hence, the FAE can generate robust features, enhancing its performance on deep feature representation learning. Several tests on three datasets show us the proper parameter settings, and the experimental results from the FAE framework and the other compared approaches demonstrate its effectiveness and robustness in terms of accuracy and elapsed time.

Recently, compressed sensing (CS) has been applied in synthetic aperture radar (SAR). A framework of mixed sparse representation (MSR) has been proposed for reconstructing SAR images due to the complicated ground features. The existing method decomposes the image into the point and smooth components, where the sparse constraint is directly applied to the smooth components. This makes it difficult to tackle the complex-valued SAR images, since the phase angles of SAR images are always stochastic. A magnitude-phase separation MSR method is proposed for CS-SAR imaging based on approximated observation. Compared to the existing method, the proposed method has better reconstruction ability, because it only imposes the sparse constraint on the magnitude of the smooth components, and therefore, the phase angles are still stochastic. Furthermore, owing to the inherent low memory requirement of approximated observation, the proposed method requires much less memory cost. In the simulation and experimental results, the proposed method deals with the complex-valued SAR images effectively and demonstrates superior performance over the chirp scaling algorithm and the existing MSR method.

As the target-radar feature is time-variant in the cognitive radar (CR) system, the target information should be continuously updated by the receiver and considered to provide the prior knowledge for the optimization of the next waveform. To solve this problem, a two-step CR waveform optimization approach for target estimation is proposed. During the first echo pulse, waveform optimization for target-radar signature estimation is done by minimizing the mean square error of target power spectral density estimation. Then, to take advantage of the temporal correlation of target scattering coefficients (TSC) during the pulses interval, a Kalman filtering-based method is used to processing successive radar echoes for TSC estimation. A convex cost function is established and the optimal solution can be obtained by the existing convex programming algorithm with multiple iterations. Finally, subject to the transmitted power, peak-to-average power ratio, and detection performance constraints, the simulation results show that the proposed waveform optimization algorithm is able to improve the performance of target estimation for extended target.

How to preserve the spectral information when enhancing the spatial details is a key issue of remote sensing image fusion. The component substitution (CS)-based fusion methods can effectively enhance the spatial details while suffering spectral distortion, and multiresolution analysis (MRA)-based methods have advantages in preserving spectral information but are not satisfactory in terms of spatial details. This paper proposes a hybrid method to integrate the advantages of CS- and MRA-based approaches. The intensity image is first obtained from an original multispectral (MS) image by hyperspherical color space (HCS) transform; then, the intensity image and original panchromatic (PAN) image are decomposed by fast discrete curvelet transform (FDCT). The focus measure operators are introduced to fused low-frequency, middle-high frequency, and fine scale subband coefficients in curvelet domain with specific fusion strategies. The final fusion image is achieved by inverse FDCT and inverse HCS transform. From the aspects of subjective and objective quality assessments, the experimental results on various types of remote sensing images including IKONOS, QuickBird, and WorldView-2 indicate that compared with existing well-known algorithms and commercial softwares, the proposed method shows obvious advantages in preserving spectral information and maintaining the spatial details.

Pan-sharpening is a procedure that fuses a low-resolution multispectral (MS) image and a high-resolution panchromatic (PAN) image into a high-resolution MS image. Existing variational pan-sharpening methods typically consider gradient as the descriptor of the spatial information and enforce this to be consistent between the PAN image and fused image because of their shared scenes. However, the gradient feature is not sufficient to exploit the abundant spatial details. Second-order derivative, which is balanced with the first-order derivative, is introduced to build a variational model such that additional geographic information is extracted from the PAN image. Then corresponding numerical schemes to determine the minimal solution of this model are provided. Through experiments on data from four different satellites, the proposed method is proven to enhance the spatial quality of the fused result and demonstrate stable performance in preserving the spectral information.

Markovian model approach with graph cuts is effectively applied for phase unwrapping (PU) in interferometric synthetic aperture radar (InSAR) domain particularly for data with big noises and discontinuities. The first-order prior model in Markov random field (MRF) is widely used in Markovian model approach, which is able to solve PU problem in homogeneous scenarios. However, the first-order MRF method cannot interpret local details well in abundant patterns of interferograms under the requirement of high accuracy due to the limitation of first-order prior information. Hence, the conventional first-order model is no longer applicable in this situation. In terms of this problem, the shortcoming of the first-order model is analyzed, and a precise second-order MRF approach is proposed in this paper. Furthermore, a graph cuts-based optimization PU algorithm is designed under a high-order MRF mode, which improves the PU accuracy in detail areas. Without apparent increase of computational complexity, the proposed method is more robust compared with the conventional first-order MRF-based methods. A set of experimental results on both simulation and real measured millimeter-wave InSAR data illustrate the effectiveness of the proposal.

Recently, object classification of remote sensing images has attracted more and more research interests due to the development of satellite and aerial vehicle technologies. Hashing learning is an efficient method to handle the huge amount of the remote sensing data. We proposed a hashing learning method named partial randomness supervised discrete hashing (PRSDH), which combines data-dependent methods and data-independent methods. It jointly learns a discrete binary codes generation and partial random constraint optimization model. By random projection, the computation complexity is reduced effectively. With the weight matrix derived from the training data, the semantic similarity between the data can be well preserved while generating the hashing codes. For the discrete constraint problem, this paper adopts the discrete cyclic coordinate descent algorithm to optimize the codes bit by bit. The experimental results show that PRSDH outperforms other comparative methods and demonstrate that PRSDH has good adaptability to the characteristic of remote sensing object.

In recent years, many deconvolution methods have been proposed for radar forward-looking super-resolution imaging based on the sparse characteristic of the targets. However, most of the deconvolution methods will be invalid due to the illposed convolution matrix under a low signal-to-noise ratio (SNR). This paper proposes a radar forward-looking super-resolution imaging method for the sparse target in the low SNR, which is based on truncated singular value decomposition (TSVD) strategy. The convolution model is reconstructed through TSVD strategy, by which the illposed character of deconvolution is modified. First, through choosing the truncated parameter in a reasonable way, the noise amplification is restrained and the main information of the target is maintained by the TSVD technique. Then, the convolution model is reconstructed based on the result of TSVD. Third, an objective function is established as the L1 constraint based on the regularization strategy. Finally, due to the fast convergence and low computational complexity, the iteratively reweighted least square method is utilized to obtain the optimal solution of the objective function. The noise amplification is suppressed while the sparse characteristic is utilized to improve the resolution. Hence, the false target is avoided and the locations of the targets are accurately recovered by the proposed method. The simulations and experimental results demonstrate that the proposed method is superior to the conventional sparse deconvolution method when the SNR is low.

Following the Kumamoto earthquake in April 2016, landslides occurred in the Minamiaso village of Japan. Classification of remotely sensed imagery has been useful for mapping of the landslide area. Many studies have focused on using object-based classification of the imagery to derive landslide maps. Also, those studies used shape or texture analysis, which requires significant time and labor, and expensive commercial software. The purpose of this research was to develop a methodology to map the landslide area rapidly and economically using object-based classification with true color satellite images. Publicly available, RapidEye images taken before and after the landslide were used to save cost. Segmentation of the imagery was completed with the mean shift algorithm in the open source noncommercial remote sensing software, Orfeo toolbox. In addition, a maximum likelihood classification tool was built in a Python framework (open source freeware) to classify the objects created from Orfeo toolbox. For initial classification of each RapidEye image for both pre- and post-landslide, the overall accuracies were 91% and 88%, and Kappa statistics were 0.89 and 0.85, respectively. The landslide area was estimated as 979,815  m² with an accuracy of 98% after human-guided refinement of the landslide features. Every procedure was implemented using open source remote sensing or geographic information system technologies, which provided cost efficiency while simultaneously providing freedom from commercial license restrictions. Also, these open source technologies provided highly flexible interoperability with a classification framework developed using Python.

Clearing large civilian areas of antipersonnel landmines is a very severe problem that should be solved efficiently. Although many methods have been developed for landmine detection and discrimination using ground penetrating radar data, the problem has not yet been properly solved, especially, as landmine and innocuous object classes are imbalanced. One-class classification is a competitive method for landmine detection as data are unbalanced, but it separates the target from outliers along the target class large variance directions, which results in higher error. As a solution, covariance-guided one-class support vector machine (COSVM) emphasizes low-variance projectional directions of the training data, which results in high accuracy and error minimization. However, in the case of a large-scale dataset, COSVM requires a large memory and enormous amount of training time. Moreover, it is inflexible with dynamic data. For these reasons, we investigate the effectiveness of incremental covariance-guided one-class support vector machine (ICOSVM) to build an adaptive landmine detection and localization system. In fact, the ICOSVM has the advantage of incrementally projecting the data onto low-variance directions, thereby improving detection performance. Experimental results have shown clearly the superiority and efficiency of the proposed landmine detection and localization system.

Continual advances in hyperspectral technology have allowed researchers to easily and rapidly estimate the copper content of Seriphidium terrae-albae, an effective indicator plant for the exploration of copper deposits. Seriphidium terrae-albae is a small semishrub with relatively small leaves. Therefore, the canopy spectra are intensely affected by the surrounding soil. This significantly affects the inversion accuracy of copper content. Spectral data transformation is important in eliminating background noise and extracting spectral features. This paper focuses on researching the effects of different spectral transformation methods on the copper content inversion accuracy in Seriphidium terrae-albae and selecting the most effective method. It begins by conducting 12 spectral transformations for original spectra and then inputting a spectrum that has been transformed into the partial least squares model (PLS) for further modeling the input parameters. The final step is selecting cross-validation R² as a standard for the most suitable spectral transformation method. The results indicate that different spectral transformation methods significantly affect the accuracy of the copper content inversion. The original spectrum that has undergone the first and second derivations can dramatically increase the copper content inversion accuracy. This includes the optimum spectral transformation as the second-order derivative of the square root of the raw spectrum. This paper also examines the variation trends of correlation between the same band and copper content under different spectral transformations. The results indicate that each band always has an optimum spectral transformation method. Therefore, this paper proposes using different spectral transformation methods at different wavelengths to form a combined spectrum within the whole wavelength. The experiments conducted as a part of this study demonstrate that the combined spectrum can greatly improve the copper content inversion accuracy in comparison with the single spectral transformation method. These findings can provide some reference for scholars, who will use hyperspectral data to estimate the metal content of plants similar to Seriphidium terrae-albae.

Hyperspectral images have the advantages of high spectral resolution and abundant spectral information, which helps researchers to analyze, process, and monitor land surface information. However, it also faces problems such as low degree of automatic data processing and poor timeliness of information extraction. This timeliness is due to the high dimensionality when extracting the target information in hyperspectral image. To solve this problem, a practical real-time hyperspectral anomaly detection system enhanced by graphics processing unit (GPU) is designed. The push-broom imaging spectrometers are adopted to obtain the image data line-by-line; then, the real-time anomaly detection algorithm based on QR decomposition is optimized using GPU (i.e., A QR decomposition is a decomposition of a matrix A into a product A = QR of an orthogonal matrix Q and an upper triangular matrix R). The results show that the parallel algorithm based on GPU is up to about 26 times faster than the serial algorithm based on central processing unit. This indicates that the proposed scheme significantly enhances the implementation efficiency while maintaining the accuracy of the results and can meet the application requirements of real-time anomaly detection well.

Mapping of peanut crops is essential in supporting peanut production, yield prediction, and commodity forecasting. While ground-based surveys can be used over small areas, the development of remote-sensing technologies could provide rapid and inexpensive crop area estimates with high accuracy over large regions. Some of these recent earth observation satellite systems, such as the Project for On-Board Autonomy Vegetation (PROBA-V), have the advantage of increased spatial and temporal resolution. With a study area located in the South Burnett region, Queensland, Australia, the primary aim of this study was to assess the ability of time-series PROBA-V 100-m normalized difference vegetation index (NDVI) for peanut crop mapping. Two datasets, i.e., PROBA-V NDVI time-series imagery and the corresponding phenological parameters generated from TIMESAT data analysis technique, were classified using maximum likelihood classification, spectral angle mapper, and minimum distance classification algorithms. The results show that among all methods used, the application of MLC in PROBA-V NDVI time series produced very good overall accuracy, i.e., 92.75%, with producer and user accuracy of each class ≥78.79  %  . For all algorithms tested, the mapping of peanut cropping areas produced satisfactory classification results, i.e., 75.95% to 100%. Our study confirmed that the use of finer resolution 100 m of PROBA-V imagery (i.e., relative to MODIS 250-m data) has contributed to the success of mapping peanut and other crops in the study area.

Several remote sensing proxies, such as image texture, have been used to describe the spatial variation of different tropical forest’s attributes at medium scales. Fourier transformed ordination (FOTO) and gray level co-occurrence matrix (GLCM) are two texture methods frequently used in these studies but seldom compared. Therefore, the objective of this study was to compare the potential of these metrics to describe and predict structural and diversity attributes of a tropical swamp forest located in southeast Mexico. FOTO and GLCM textures were extracted from the panchromatic (Pan) band, the near-infrared (NIR) and red (R), of a very high spatial resolution (VHSR) image. Six different types of models that incorporated between one and three texture variables were constructed to describe each forest attribute. Simpson (R²  =  0.71) and Shannon (R²  =  0.67) diversity indices followed by the aboveground biomass (R²  =  0.65) were the attributes that obtained models with a higher goodness-of-fit values. In terms of its modeling potential, GLCM greatly surpassed FOTO textures. Our results show that VHSR textures (especially R- and NIR-derived GLCM metrics) allow for a reliable estimation of important tropical swamp forest attributes associated with its diversity and structure.

Bayesian network classifiers (BNCs) are now among the most used supervised probabilistic methods for remote sensing image classification. Our contribution lies in two principal points. First, the investigation of the applicability of Kruskal’s algorithm constructs the optimal tree structure of multinet Bayeisan network classifier (MBNC). Second, the focus on MBNC’s advantages is over other classical BNCs, such as naive Bayes classifier (NBC), tree augmented naive Bayes classifier (TANC), forest augmented naive Bayes classifier (FANC), and state-of-the art competitor classifiers such as maximum likelihood classifier (MLC) and support vector machine (SVM) classifier. While classical BNCs have a mere network for all predefined classes, MBNC has as many local Bayesian networks as the predefined classes. Hence, through a statistical evaluation and a visual inspection, our objective is to emphasize the contribution of MBNC to enhance the accuracy of urban land cover map obtained by classification of remotely sensed image. Performances of developed BNCs (NBC, TANC, FANC, and MBNC) are experimentally assessed using a multispectral image acquired on July 12, 2010, by Alsat-2A Algerian satellite. Based on a confusion matrix, overall accuracy, and Kappa statistic, results indicate that MBNC largely outperforms classical BNCs (NBC, TANC, and FANC) and probabilistic MLC, but performs slightly better than an SVM classifier. Due to its specific-class local network, MBNC gives a powerful tool for a better discrimination between different correlated spectral classes.

This work was supported by the National Center for Space Studies in the framework of the future mission surface water and ocean topography (SWOT). The estuarine zones with their complex hydrodynamics represent a great challenge for the future SWOT mission. The Seine estuary was selected by the SWOT Science Definition Team for the calibration/validation of the mission. Thus the aim of this study is to investigate the temporal hydrodynamic variability in the Seine estuary and the SWOT ability to reproduce this variability and hydrodynamic phenomena. For this, we used several statistical techniques and frequency analyses applied on the tide-gauge measurements of the Seine estuary and on SWOT simulated data. SWOT samples (SWOT simulated data) have been extracted synthetically from observations using the number of overpasses per repeat cycle [21 days (d)] in order to study the ability of SWOT to reproduce the time variability of the hydrodynamics. Results have demonstrated similar frequencies of multiyear, annual, and intraannual along the estuary from the Seine discharge to the Cherbourg sea level. The 11- to 22-d frequency has been only observed for estuarine and coastal time series suggesting the tide effect in these contexts. The distribution of levels has shown a high influence of tides and discharges at the downstream and the upstream, respectively. Low frequencies of water level are explained by the climate patterns of North Atlantic Oscillation with a mean range of 65%, whereas the high frequencies are associated to the periods of flooding and stormy events. Finally, simulated SWOT samples of water level show a better restitution of observations for high overpasses number. The results have shown a better restitution of hydrodynamic variability by SWOT in the river and upstream of estuary than in the downstream of estuary and coastal zone: in the estuary downstream and the coast, the annual mode is underexpressed while the 3-month (m) mode is overexpressed, which means the annual mode is underestimated and the 3-m mode is amplified by SWOT.

The regional surface soil heat flux (G₀) is very important for the large-scale land surface process modeling, and particularly for the regional evapotranspiration estimation. Harmonic analysis method (HM) could obtain regional G₀ using remote sensing data. However, it has been demonstrated that HM cannot work well during nighttime due to the nonharmonic nature of G₀. Encouraged by the half-order method (HO), which can resolve nonharmonic series, this study combined the two methods of HM and HO in order to obtain an accurate estimation of G₀ for full day. The combined method was assessed with in-situ measurements and was applied to the Heihe River Basin (HRB) of northwest China using multiremote sensing observations at 30-min intervals. The below-canopy soil surface temperature (T_s) is one of the key variables in the model, which can be estimated from canopy temperature (T_c). The relationship of T_c and T_s was discussed. Results show that the nighttime G₀ improves with coefficient of determination (R²) from 0.17 to 0.66 by HO method over maize and from 0.09 to 0.81 over grassland, and root mean square error decreases from 24.5 to 16.6  W m^−  2, 16.1 to 8.68  W m^−  2 over maize and grassland, respectively. Regional G₀ for bare soil is larger than the vegetation surface, and values in the middle reach area of HRB are generally higher than the upper reach area because of larger fractional vegetation cover in the upper reach. For maize, grassland, and vegetable surfaces (all are dense vegetation in July in this study), the combined method performed better when using T_c (almost coincides with remote sensing measurement land surface temperature) than estimated T_s by a simple empirical formula provided by Murray and Verhoef. The combined method gets more accurate diurnal G₀ than individual HM. However, a more theoretical relationship of T_c and T_s is required when more accurate G₀ is needed in other studies.

This paper considers the problem of block-sparse recovery for two-dimensional inverse synthetic aperture radar (ISAR) imaging. According to the scatterer distribution of the target scene in ISAR image, the continuity pattern in both range and cross-range domains should be considered. Therefore, the sparsity of each grid cell is controlled by four neighboring hyperparameters and the relevance between neighboring coefficients is determined by coupling parameters, which are data-dependent, so the estimation is done adaptively by an expectation–maximization algorithm. To model the pattern dependencies among neighboring scatterers on range-Doppler domain, we develop the data-dependent coupling parameters method to capture continuity pattern of ISAR signals. Simulation results show that the proposed method can achieve improvement in terms of entropy, image entropy, and image contrast. Moreover, our algorithm effectively improves reconstruction of target scene in noiseless and noisy case compared with other methods.

Desert locusts have attacked crops since antiquity. To prevent or mitigate its effects on local communities, it is necessary to precisely locate its breeding areas. Previous works have relied on precipitation and vegetation index datasets obtained by satellite remote sensing. However, these products present some limitations in arid or semiarid environments. We have explored a parameter: soil moisture (SM); and examined its influence on the desert locust wingless juveniles. We have used two machine learning algorithms (generalized linear model and random forest) to evaluate the link between hopper presences and SM conditions under different time scenarios. RF obtained the best model performance with very good validation results according to the true skill statistic and receiver operating characteristic curve statistics. It was found that an area becomes suitable for breeding when the minimum SM values are over 0.07  m3  /  m3 during 6 days or more. These results demonstrate the possibility to identify breeding areas in Mauritania by means of SM, and the suitability of ESA CCI SM product to complement or substitute current monitoring techniques based on precipitation datasets.

The alignment of images acquired using the synthetic aperture radar (SAR) is a central task in image processing for remote sensing. Owing to speckle noise and the particular features of SAR imaging, stable feature detection and accurate matching remain challenging tasks. We propose a registration method for SAR images. In feature detection, we develop a hybrid feature detection method to identify structural and textural features. A stable convex corner point feature is detected using stable convex polygons obtained by a proposed method for local stable extremal region extraction and convex polygon fitting. The stable local extremum points in each convex polygon are detected by an improved scale invariant feature transform method. In feature matching, a multifeature constraint matching method is designed for accurate matching. Coarse matching is implemented using the constraints on the region and its shape as well as the network. The use of stable local extremum points with spatial constraints can eliminate mismatches and yield a fine match. The results of experiments verified the effectiveness and accuracy of the proposed method.

Cyanobacterial blooms are a nuisance and a potential hazard in freshwater systems worldwide. Remote sensing has been used to detect cyanobacterial blooms, but few studies have distinguished among genera of cyanobacteria. Because some genera are more likely to be toxic than others, this is a useful distinction. Hyperspectral imaging reflectance microscopy was used to examine cyanobacteria from Upper Klamath Lake, Oregon, at high spatial and spectral resolution to determine if two species found commonly in the lake, Aphanizomenon flos-aquae and Microcystis aeruginosa, can be separated spectrally. Of the analytical methods applied, a spectral shape algorithm applied to the derivative was found to be most successful in classifying these species in microscope scenes. Further work is required to determine if the spectral characterization of cyanobacterial genera can be scaled up to remote sensing applications.

Several quasi-analytical algorithm (QAA) versions were developed to make it suitable for different inland water systems. QAABBHR and QAAOMW were reparameterized based on two reservoirs from the Tietê River cascading system (São Paulo State, Brazil), which present widely differing compositions. Considering the purpose of monitoring the entire cascade through a unique QAA version, we aimed to assess the suitability of these two QAA versions and, in addition, another two QAA native forms (versions 5 and 6), for retrieving inherent optical properties (IOPs) in Ibitinga hydroelectric reservoir (IHR), situated in the same cascading system. In addition to that, we addressed bio-optical characterization of IHR, using spectral and water quality data collected in a field campaign conducted in July 2016. Wide spatial variability of optically significant constituent (OSC) in IHR and colored dissolved organic matter predominance in its absorption budget was observed. None of the tested QAA versions were completely suitable in retrieving absorption coefficients for IHR in all wavelengths. However, results for wavelengths commonly used as proxy for OSC concentration retrieval were satisfactory in some of the models. Therefore, the results obtained in this study shows that QAAs versions can be used for specific purposes (e.g., chlorophyll-a mapping), by employing the best model for IOPs retrieval at a specific wavelength. This highlights the challenge of copying with high optical variability in cascading systems. In this sense, further research is necessary, for either achieving a QAA reparameterized version appropriate for aquatic systems with widely differing optical properties or another analytical scheme.

Natural oil and gas are important sources of energy worldwide and their exploration and exploitation have significantly increased due to the global demand. The transportation of these valuable resources greatly depends on pipelines; however, pipeline leakages have huge economic and environmental impacts warranting an effective operational methodology for pipeline monitoring. We proposed a method for mapping soil contamination due to pipeline leakage in Dixonville, Alberta, Canada. In particular, very high-resolution unmanned aerial vehicle (UAV) imagery and electromagnetic induction (EM) surveying data were analyzed using a hierarchical object-based random forest (RF) algorithm. In level-1 classification, a land cover map was produced using UAV data. Next, all land cover classes, excluding contaminated soil, were masked out. In level-2 classification, the contaminated soil class was further partitioned into three subclasses representing varying degrees of contamination. Specifically, we proposed a salinity index, named the normalized salinity index, to detect areas of soil contamination. The salinity index proposed herein, as well as several other salinity indices and UAV bands, were used as input features for level-2 classification. An overall classification accuracy of about 77% was achieved for level-2 classification using the proposed method. The results demonstrate that the synergistic use of high spatial resolution UAV imagery and EM data is very promising for detecting soil contamination and examining ecosystem disturbance due to pipeline leakage.

Population research on the saguaro cactus (Carnegiea gigantea) has historically been limited to repeat monitoring efforts in small plots with areas of 1 to 4 hectares. Large extent saguaro censuses of areas >2.59  km² have not been undertaken. This research developed an automated shadow detection method for mapping mature saguaros in digital aerial imagery using saguaro shadows as proxies for saguaro locations. The shadow method detected 446,092 saguaro shadows in aerial imagery of 231  km² of the Saguaro National Park (SNP) near Tucson, Arizona. The results were validated against saguaro location data in 11 small plots provided by SNP staff. The shadow method correctly identified 58% of mature saguaros in these 11 plots. Twenty-one percent of the saguaros in the plots did not produce a shadow signature in the imagery. Fieldwork was conducted in two plot locations to investigate reasons why these mature saguaros failed to produce shadow signatures in the aerial imagery, which showed that vegetation interference, rocky slopes, and shadow alignments were the primary factors that precluded shadow signatures from appearing in the imagery. This research provides a methodology for automated large-area saguaro mapping, which will be useful in long-term population monitoring and saguaro ecology research.

A semianalytical model developed to estimate the Secchi disk depth (ZSD) was used in eutrophic-to-hypereutrophic reservoirs (Ibitinga, Ibi, and Barra Bonita, BB) placed in the cascade system of the Tietê River, Brazil. The model was evaluated using the simulated remote sensing reflectance based on the Ocean and Land Color Instrument/Sentinel-3A and the Operational Land Imager/Landsat-8 from both reservoirs. Three quasianalytical algorithm (QAA) versions (QAAv5, QAAM14, and QAAW16) were evaluated to derive the absorption and backscattering coefficients, and then used for ZSD retrieval. For BB, where the chlorophyll-a concentration exceeded 200  mg m  −  3, the model based on QAAv5 showed high uncertainties while the QAAW16, which was originally parameterized for BB showed better performance regarding the ZSD retrieval (mean absolute percentage errors—MAPE of 22%). However, QAAW16 did not perform satisfactorily for Ibi, which is dominated by colored dissolved organic matter (CDOM). For Ibi, QAAv5 provided the best result with MAPE of 34.60%, followed by QAAM14 with 34.65%. QAA-based ZSD models tend to perform poorly in waters with high concentration of chlorophyll-a possibly due to phytoplankton package effect, whereas the same models may require additional parameterization in waters dominated by CDOM. Landsat-8 data showed significant potential for ZSD retrieval in inland waters.

Effectively assessing cadmium (Cd) contamination in crops is crucial for the sustainable development of an agricultural ecosystem and for environmental security. We developed an integrated stress index (SI) based on two phenological metrics to effectively evaluate Cd stress in rice crops. The selected four experimental areas are located in Zhuzhou City, Hunan Province, China. Six Sentinel-2 images were acquired in 2017, and heavy metal concentrations in soil were measured. The change rate of CIre (CRCIre) and the time-integrated CIre (TICIre) were obtained from daily red-edge chlorophyll index (CIre) time-series using Sentinel-2 data. The CRCIre and TICIre were used to characterize the photosynthetic rate and biomass, respectively. SI was calculated by Fisher discriminant analysis based on CRCIre and TICIre from two experimental areas, and it was verified using another two experimental areas. The results were the following: (i) when SI  ≥  0, rice was under mild stress and when SI  <  0, rice was under severe stress. (ii) The SI effectively evaluated Cd stress levels with an overall discriminatory accuracy of 86.02%. This research provides a potential new method to evaluate Cd stress in rice by remote sensing through phenology.

Heavy-metal pollution of crops seriously affects food safety and human health in China. The purpose of this study is to estimate the concentrations of heavy metals economically and accurately using spectroscopy. The spectral data of winter wheat canopy at 61 sampling points were collected in Longkou City, and the concentration of Pb, Zn, Cd, Cr, Ni, and Hg was measured. Eight estimation models were established based on the back-propagation neural network (BPNN), the partial least squares regression, and the stepwise multiple linear regression methods combined with reflectance spectra, first-order derivative of spectral reflectance (FDR), second-order derivative of spectral reflectance, and spectral parameters (SPs). Spatial interpolation was used to map the heavy-metal concentration. The results showed that the best model for Pb, Zn, and Cd was the BPNN model with SPs, for Cr, Ni, and Hg was the BPNN model with FDR. The concentration of Hg was lower in the northwest, Pb and Zn were higher in the middle and south, Cd was lower in the middle and west, and Cr and Ni were higher in the southeast. This study will provide effective technical support for the economical and accurate monitoring of heavy metals.

Quantification of the spatial distribution of fractional cover of plant functional types is important for peatland ecosystems analysis. Using a crisp classification to produce a map of the spatial patterns of plant functional types in peatlands is difficult due to their complex and heterogeneous characteristics. We applied multiple endmember spectral mixture analysis (MESMA) on Airborne Imaging Spectrometer for Applications Eagle data to generate fractional covers of peatland plant functional types and further to assess the accuracy of soft classification at the subpixel scale. This was achieved by constructing the spectral library and selecting the optimal endmembers models (two-, three-, and four-endmember MESMA models) to derive the peatland plant functional types (shrub, graminoid, moss, and forbs) fractional covers. The final fractional cover pattern in each case was produced by joining the results from the multiendmembers models as judged by the root mean squared error thresholds to select the best-fit model for each pixel by minimization of pixel-scale fraction errors. Based on the thresholds, the three- and four-endmember MESMA models and their combination were further tested for their performance using an independent in situ dataset of the four-plant functional types. Their accuracy performance was quantified by a subpixel confusion matrix and residual analysis. The results showed that the combined MESMA model was most accurate in deriving the patterns of fractional cover for shrub, graminoid, and moss compared with the performance of the three- and four-endmember MESMA models. However, all models performed poorly in predicting the fractional cover of the sparsely distributed forbs. The results demonstrate the superiority of MESMA models in extracting the patterns of different plant functional types in peatlands as judged by the accuracy of subpixel fractional extraction.

To overcome the illposed problems existing in small baseline subsets deformation, the Liu-type estimator is used to improve the accuracy of calculation. The Liu-type estimator introduces parameter d to improve the fitting and statistical properties after the condition number of the design matrix is reduced to the desired level, which makes the solution of illposed problems robust. In the experiment, subsidence rates relevant to the Beijing area are computed from the ENVISAT satellite radar data from the European Space Agency from 2007 to 2010. The experimental results show that the accuracy of the deformation rates computed with the Liu-type estimator is clearly improved.