This PDF file contains the front matter associated with SPIE Proceedings Volume 9121 including the Title Page, Copyright information, Table of Contents, Introduction, and Conference Committee listing.

The Mahalanobis Taguchi System (MTS) is a relatively new tool in the vehicle health maintenance domain, but has some distinct advantages in current multi-sensor implementations. The use of Mahalanobis Spaces (MS) allows the algorithm to identify characteristics of sensor signals to identify behaviors in machines. MTS is extremely powerful with the caveat that the correct variables are selected to form the MS. In this research work, 56 sensors monitor various aspects of the vehicles. Typically, using the MTS process, identification of useful variables is preceded by validation of the measurements scale. However, the MTS approach doesn’t directly include any mitigating steps should the measurement scale not be validated. Existing work has performed outlier removal in construction of the MS, which can lead to better validation. In our approach, we modify the outlier removal process with more liberal definitions of outliers to better identify variables’ impact prior to identification of useful variables. This subtle change substantially lowered the false positive rate due to the fact that additional variables were retained. Traditional MTS approaches identify useful variables only to the extent they provide usefulness in identifying the positive (abnormal) condition. The impact of removing false negatives is not included. Initial results show our approach can reduce false positive values while still maintaining complete fault identification for this vehicle data set.

Bayesian statistics offers a well-founded and powerful fusion methodology also for the fusion of heterogeneous information sources. However, except in special cases, the needed posterior distribution is not analytically derivable. As consequence, Bayesian fusion may cause unacceptably high computational and storage costs in practice. Local Bayesian fusion approaches aim at reducing the complexity of the Bayesian fusion methodology significantly. This is done by concentrating the actual Bayesian fusion on the potentially most task relevant parts of the domain of the Properties of Interest. Our research on these approaches is motivated by an analogy to criminal investigations where criminalists pursue clues also only locally. This publication follows previous publications on a special local Bayesian fusion technique called focussed Bayesian fusion. Here, the actual calculation of the posterior distribution gets completely restricted to a suitably chosen local context. By this, the global posterior distribution is not completely determined. Strategies for using the results of a focussed Bayesian analysis appropriately are needed. In this publication, we primarily contrast different ways of embedding the results of focussed Bayesian fusion explicitly into a global context. To obtain a unique global posterior distribution, we analyze the application of the Maximum Entropy Principle that has been shown to be successfully applicable in metrology and in different other areas. To address the special need for making further decisions subsequently to the actual fusion task, we further analyze criteria for decision making under partial information.

In typical classification problems the data used to train a model for each class is often correctly labeled, and so that fully supervised learning can be utilized. For example, many illustrative labeled data sets can be found at sources such as the UCI Repository for Machine Learning (http://archive.ics.uci.edu/ml/), or at the Keel Data Set Repository (http://www.keel.es). However, increasingly many real world classification problems involve data that contain both labeled and unlabeled samples. In the latter case, the data samples are assumed to be missing all class label information, and when used as training data these samples are considered to be of unknown origin (i.e., to the learning system, actual class membership is completely unknown). Typically, when presented with a classification problem containing both labeled and unlabeled training samples, a technique that is often used is to throw out the unlabeled data. In other words, the unlabeled data are not included with existing labeled data for learning, and which can result in a poorly trained classifier that does not reach its full performance potential. In most cases, the primary reason that unlabeled data are not often used for training is that, and depending on the classifier, the correct optimal model for semi-supervised classification (i.e., a classifier that learns class membership using both labeled and unlabeled samples) can be far too complicated to develop. In previous work, results were shown based on the fusion of binary classifiers to improve performance in multiclass classification problems. In this case, Bayesian methods were used to fuse binary classifier fusion outputs, while selecting the most relevant classifier pairs to improve the overall classifier decision space. Here, this work is extended by developing new algorithms for improving semi-supervised classification performance. Results are demonstrated with real data form the UCI and Keel Repositories.

The Process for Probabilistic Multi-Source Multi-INT Intel Fusion Benefit Analysis** (PIFBA) provides a methodology for statistically computing the probability of detecting, locating, and tracking objects in order to assess current capabilities or the benefit of adding or removing ISR assets in order to obtain the statistically optimal result. The PIFBA process defines the approach to calculate the probabilistic benefits or rewards associated with integrating or fusing multisource multi-INT products across a wide range of platforms, sensors, environmental conditions and target objects. This process applies to analyzing ISR capabilities, effectiveness, and gaps, as well as the benefits of applying existing or new technology and tactics. The PIFBA process was designed to answer the following questions:  Based on the defined ISR assets – what is the probability that we know a piece of Intel with sufficient accuracy and timeliness to be of value to the analyst or warfighter?  What is the benefit of adding new data via additional platforms, sensors, or processing?  What is the benefit of adding new systems or technology and to what degree of performance must they exhibit in order to affect the statistical outcome?

In this work, simple, generic models of chemical sensing are used to simulate sensor array data and to illustrate the impact on overall system performance that specific design choices impart. The ability of multisensor systems to perform multianalyte detection (i.e., distinguish multiple targets) is explored by examining the distinction between fundamental design-related limitations stemming from mismatching of mixture composition to fused sensor measurement spaces, and limitations that arise from measurement uncertainty. Insight on the limits and potential of sensor fusion to robustly address detection tasks in realistic field conditions can be gained through an examination of a) the underlying geometry of both the composition space of sources one hopes to elucidate and the measurement space a fused sensor system is capable of generating, and b) the informational impact of uncertainty on both of these spaces. For instance, what is the potential impact on sensor fusion in an open world scenario where unknown interferants may contaminate target signals? Under complex and dynamic backgrounds, decision rules may implicitly become non-optimal and adding sensors may increase the amount of conflicting information observed. This suggests that the manner in which a decision rule handles sensor conflict can be critical in leveraging sensor fusion for effective open world sensing, and becomes exponentially more important as more sensors are added. Results and design considerations for handling conflicting evidence in Bayes and Dempster-Shafer fusion frameworks are presented. Bayesian decision theory is used to provide an upper limit on detector performance of simulated sensor systems.

As computer network security threats increase, many organizations implement multiple Network Intrusion Detection Systems (NIDS) to maximize the likelihood of intrusion detection and provide a comprehensive understanding of intrusion activities. However, NIDS trigger a massive number of alerts on a daily basis. This can be overwhelming for computer network security analysts since it is a slow and tedious process to manually analyse each alert produced. Thus, automated and intelligent clustering of alerts is important to reveal the structural correlation of events by grouping alerts with common features. As the nature of computer network attacks, and therefore alerts, is not known in advance, unsupervised alert clustering is a promising approach to achieve this goal. We propose a joint optimization technique for feature selection and clustering to aggregate similar alerts and to reduce the number of alerts that analysts have to handle individually. More precisely, each identified feature is assigned a binary value, which reflects the feature's saliency. This value is treated as a hidden variable and incorporated into a likelihood function for clustering. Since computing the optimal solution of the likelihood function directly is analytically intractable, we use the Expectation-Maximisation (EM) algorithm to iteratively update the hidden variable and use it to maximize the expected likelihood. Our empirical results, using a labelled Defense Advanced Research Projects Agency (DARPA) 2000 reference dataset, show that the proposed method gives better results than the EM clustering without feature selection in terms of the clustering accuracy.

The Automatic Identification System (AIS) is a self-reporting system based on VHF radio to transmit a vessel’s identity, position, speed, heading and other parameters to improve maritime domain awareness. However, AIS information can be programmatically spoofed by terrorists or other criminals, who often choose to masquerade as innocent civilians and exploit the vulnerabilities of military and civilian infrastructures for their purposes. Therefore, detecting and localizing a spoofing AIS ship become a critical and challenging issue for maritime security. This paper presents an algorithm to detect and geolocalize a spoofing AIS emitter using space-based AIS signals with its Doppler frequency. With an AIS signal sensor on a fast orbiting satellite, the measured AIS Doppler frequency of an AIS emitter can be used to define a double-napped cone of which the satellite is at its vertex and satellite velocity coincides with its axis, such that the theoretical Doppler frequency derived from the radial velocity to the AIS emitter matches the measured Doppler frequency. All such matches can only lie on either cone extending out from the satellite, which cuts the Earth’s surface in two curves, so we know that the AIS emitter must lie somewhere on these curves. Two such AIS Doppler frequency measurements for the same stationary AIS emitter produce two valid curves which intersect at the position of the AIS emitter. Multiple Doppler frequency measurements can be used to better estimate the position fix of an AIS emitter, hence determine the spoofing AIS ship if the estimated position fix unreasonably differs from the position carried in its AIS message. A set of formulas are derived which relate an AIS emitter position to its Doppler frequency measurements.

This paper reports a metamaterial inspired combined inductive-capacitive sensing method for detecting and distinguishing metallic and non-metallic objects. Metallic and non-metallic objects can be distinguished by measuring both of their inductive and capacitive responses based on the fact that they respond differently to inductive and capacitive sensing. The proposed method is inspired by metamaterial structures. Both inductive and capacitive sensing are simultaneously realized when the sensor is operating at off-resonant frequencies. The proposed method is demonstrated with typical printed circuit board (PCB) technology. The designed sensor can distinguish the metallic and dielectric objects with a sensing range about 10 mm, showing a competitive performance compared with commercially available proximity sensors.

The acquisition of synchronous EO imagery and gamma radiation data in aerial overflights of an unmanned aircraft can provide valuable spatial context for radioactive source mapping. Using image-based 3D reconstruction methods, a terrain map can be generated and used to reason about more likely radiation source locations. For instance, vehicles may be likely hiding places for nuclear materials, so a source model with assigned probability is used at the vehicle to reduce the overall uncertainty in position estimation. Environment reconstructions based on EO imagery with a mapped gamma radiation overlay provide intrinsic correlations between the datasets. Using radioactive material dispersion models or point source models, the derived correlations serve to enhance coarse gamma radiation data. The use of autonomous unmanned aircraft provide a valuable tool in acquiring these data as they are capable of accurate and repeatable position control while eliminating exposure danger to the operators. In this experiment, two sources (.084 Ci ¹³⁷Ce and .00048 Ci ¹³³Ba) were distributed in a field with varying terrain and a scan was conducted using the Virginia Tech Yamaha RMAX autonomous helicopter equipped with a two-camera imaging system and a NaI scintillation-type spectrometer. Terrain reconstruction was conducted using both structure from motion (SfM) and stereo vision techniques, and radiation data synchronized to the imagery was overlaid.

Intelligent transportation system (ITS) applications are expected to provide a more efficient, effective, reliable, and safe driving experience, which can minimize road traffic congestion resulting in a better traffic flow management. To efficiently manage traffic flows, in this paper, we compare the effectiveness of two well-known vehicle routing algorithms: the Dijkstra's shortest path algorithm and the A* (Astar) algorithm in terms of the total travel time and the travel distance. To this end, we built a generic ITS test-bed and created several real-world driving scenarios using field and simulation data to evaluate the performance of these two routing algorithms. The dataset used in our simulation is six weeks traffic volume data from 08/01/2012 to 09/27/2012 in the Maryland (MD)/Washington DC and Virginia (VA) area. Our simulation data shows that an increase in network size results in scalability problems as the efficiency and effectiveness of these algorithms diminishes in larger road networks with greater traffic volume densities, flow rates, and congested conditions. In addition, the imprecision of the road network increases as the network size and the traffic volume density increases. Our study shows that the ability of these vehicular routing algorithms to adaptively route traffic depends on the size and type of road networks, and the current roadway conditions.

As new imaging modalities arise, the problem of inpainting becomes increasing important. Typical techniques for inpainting are completely determined by the penalization term used in the optimization scheme. These methods range from minimizing over total variation to finding a sparsest solution in a given basis to minimizing the Ginzburg-Landau energy. In this paper, we propose a novel approach to inpainting of remote sensing images, which uses previous measurements taken from heterogeneous image soures in conjunction with these well studied penalization methods. These previous measurements could be images with different illumination or weather conditions, images with spatio-temporal changes, or even all together different imaging modalities. Our approach utilizes manifold learning techniques such as diffusion maps or Laplacian eigenmaps that are applied to each image. This is followed by learning a rotation between the two feature spaces in an effort to place data points from both images in a common feature space. Then, we apply a novel preimage algorithm to the fused data in conjunction with an inpainting penalization method to recreate the missing pixels.

The study was conducted in an oil producing environment dominated by mangrove and swamp vegetation in Niger Delta, Nigeria. Ancillary data including oil pipeline map and GPS of spill points were used in selecting sample sites to identify and detect polluted locations. A number of polluted and non-polluted sites were selected and vegetation spectral reflectance and indices for these sample sites were extracted from TM data of January and December 1986. A statistical T-test was used to test for significant differences between vegetation spectral reflectance and indices from polluted and non-polluted sites. The initial results from the analysis of spectral reflectance between polluted and non-polluted did not show any significant difference in all the six spectral bands with p-value <0.005. The results from analysis of various vegetation indices some did not show any significance differences between the polluted and non-polluted sites (e.g. the SRI, SAVI and EVI2). Other VIs (NDVI, MSAVI2 and ARVI2) showed significant differences between the polluted and non-polluted sites. From these preliminary results we can conclude that pollution from oil spills may result to the changes in leaf biochemistry of the Mangroves in the Niger Delta which are detectable from remote sensing data. Future work will focus on undertaking further temporal analysis of additional spill sites to determine what quantity of spilt oil arises in spectral changes of vegetation.

We presented a system to display nightime imagery with natural colors using a public database of images. We initially combined two spectral bands of images, thermal and visible, to enhance night vision imagery, however the fused image gave an unnatural color appearance. Therefore, a color transfer based on look-up table (LUT) was used to replace the false color appearance with a colormap derived from a daytime reference image obtained from a public database using the GPS coordinates of the vehicle. Because of the computational demand in deriving the colormap from the reference image, we created an additional local database of colormaps. Reference images from the public database were compared to a compact local database to retrieve one of a limited number of colormaps that represented several driving environments. Each colormap in the local database was stored with an image from which it was derived. To retrieve a colormap, we compared the histogram of the fused image with histograms of images in the local database. The colormaps of the best match was then used for the fused image. Continuously selecting and applying colormaps using this approach offered a convenient way to color night vision imagery.

Researchers at the University of Alaska Anchorage and University of Colorado Boulder have built a low cost high performance and efficiency drop-in-place Computational Photometer (CP) to test in field applications ranging from port security and safety monitoring to environmental compliance monitoring and surveying. The CP integrates off-the-shelf visible spectrum cameras with near to long wavelength infrared detectors and high resolution digital snapshots in a single device. The proof of concept combines three or more detectors into a single multichannel imaging system that can time correlate read-out, capture, and image process all of the channels concurrently with high performance and energy efficiency. The dual-channel continuous read-out is combined with a third high definition digital snapshot capability and has been designed using an FPGA (Field Programmable Gate Array) to capture, decimate, down-convert, re-encode, and transform images from two standard definition CCD (Charge Coupled Device) cameras at 30Hz. The continuous stereo vision can be time correlated to megapixel high definition snapshots. This proof of concept has been fabricated as a fourlayer PCB (Printed Circuit Board) suitable for use in education and research for low cost high efficiency field monitoring applications that need multispectral and three dimensional imaging capabilities. Initial testing is in progress and includes field testing in ports, potential test flights in un-manned aerial systems, and future planned missions to image harsh environments in the arctic including volcanic plumes, ice formation, and arctic marine life.

We present a camera system for instantaneous, non-destructive capture of spectral signatures for forensic analysis. Our system detects highly probative samples in the forensic scene mixed by the multiple target objects by combining a coded aperture snapshot spectral imager with a multi-spectral detection algorithm. An Adaptive Cosine Estimator (ACE) is used to quantitatively detect and classify the probative samples from the decoded spectral datacube. In this paper, we demonstrate selected results using our system for luminescence characteristics and spectral classification of a number of samples.

A plane wave is scattered by a potential of bounded support. Translation, rotation and reflection of the potential, q0 induce transformations of the scattered wave. The latter can be represented by means of Born sequences, where q0 appears under the integral sign: non-local formulas are thus derived, the properties of which are discussed. Next, the symmetries induced by the 1^st BORN approximation are addressed. Invariance of the squared modulus of the scattering amplitude holds for translation and reflection. The transformation T_ε := 1₃ +Σ³ℓ=1ε_ℓA_ℓ, with {ε_ℓ;} real and {A_ℓ} the generators of rotations in IR³, is investigated. Conditions on the {ε _ℓ} are derived, by which the scattering amplitude coming from the first BORN approximation is invariant to T_ε. As an application, these “false symmetries” are compared to those induced by limited angular resolution of a detector in light scattering experiments. Namely, scattering patterns are made available by the TAOS (Two-dimensional Angle-resolved Optical Scattering) method, which consists of detecting single airborne aerosol particles and collecting the intensity of the light they scatter from a pulsed, monochromatic laser beam. The optics and the detector properties determine the resolution at which a pattern is saved. The implications on the performance of TAOS pattern analysis are briefly discussed.

In this paper, we propose a unified framework for color filter array (CFA) interpolation and visible/NIR image combination. The proposed method aims to reconstruct a high quality image from raw CFA data and the corresponding NIR image, similar to a multi-spectral fusion of color and NIR images. Based on NIR image, we impose a sparse constraint of gradient difference to modify the traditional color interpolation. The experiments indicate the effectiveness of our hybrid scheme to acquire joint color and NIR information in real-time, and show that this hybrid process can generate a better color image when compared with treating interpolation and fusion separately.

A mobile robot operating in a netcentric environment can utilize offboard resources on the network to improve its local perception. One such offboard resource is a world model built and maintained by other sensor systems. In this paper we present results from research into improving the performance of Deformable Parts Model object detection algorithms by using an offboard 3D world model. Experiments were run for detecting both people and cars in 2D photographs taken in an urban environment. After generating candidate object detections, a 3D world model built from airborne Light Detection and Ranging (LIDAR) and aerial photographs was used to filter out false alarm using several types of geometric reasoning. Comparison of the baseline detection performance to the performance after false alarm filtering showed a significant decrease in false alarms for a given probability of detection.

We address the problem of fusing laser and RGB-Data from multiple robots operating in close proximity to one another. By having a team of robots working together, a large area can be scanned quickly, or a smaller area scanned in greater detail. However, a key aspect of this problem is the elimination of the spurious readings due to the robots operating in close proximity. While there is an extensive literature on the mapping and localization aspect of this problem, our problem differs from the dynamic map problem in that it involves at one kind of transient map feature, robots viewing other robots, and we know that we wish to completely eliminate all such mutual views. In prior work, we investigated the problem of fusing laser data from multiple robots in such a manner as to reject this spurious data from other robots. This work showed that a combination of local robot-based direction filtering and global map-based visibility filtering at a central map server removed 91% of the spurious data and resulted in a 98% quality improvement. In this paper we additionally consider the problem of fusing RGB-D data generated by a stereo-camera sensor. An approach based on a model of human visual attention is presented and compared with our prior work and with other related work. This approach is an order of magnitude faster than the prior work and yet rejects 73% of the spurious data producing a 55% quality improvement. Results are shown for this approach for two experiments with a two robot team operating in a confined indoor environment (4m x 4m).

This paper discusses selected aspects of our effort to implement rudimentary emulations of brain regions and their synaptic connectivity (a paradigm we refer to as neurobiomimetic), including in particular the Neurobiomimetic Cognitive Architecture Foundation Framework (NCAFF) we developed. Different instances of neurobiomimetic emulations are possible and we refer to those instances as animats. NCAFF addresses the overwhelming developmental challenge of implementing animats, separating animat-development process from the lower-level details. Approaches such as NCAFF enable feasible building of animats which, by emulating aspects of brain structure and processes, may constitute a particularly promising path to cognitive systems.

An integrated multi-sensor fusion framework for localization and mapping for autonomous navigation in unstructured outdoor environments based on extended Kalman filters (EKF) is presented. The sensors for localization include an inertial measurement unit, a GPS, a fiber optic gyroscope, and wheel odometry. Additionally a 3D LIDAR is used for simultaneous localization and mapping (SLAM). A 3D map is built while concurrently a localization in a so far established 2D map is estimated with the current scan of the LIDAR. Despite of longer run-time of the SLAM algorithm compared to the EKF update, a high update rate is still guaranteed by sophisticatedly joining and synchronizing two parallel localization estimators.

Future smart manufacturing systems will include more complex coordination of mobile manipulators (i.e., robot arms mounted on mobile bases). The National Institute of Standards and Technology (NIST) conducts research on the safety and performance of multiple collaborating robots using a mobile platform, an automatic guided vehicle (AGV) with an onboard manipulator. Safety standards for robots and industrial vehicles each mandate their failsafe control, but there is little overlap between the standards that can be relied on when the two systems are combined and their independent controllers make collaborative decisions for safe movement. This paper briefly discusses previously uncovered gaps between AGV and manipulator standards and details decision sharing for when manipulators and AGVs are combined into a collaborative, mobile manipulator system. Tests using the NIST mobile manipulator with various control methods were performed and are described along with test results and plans for further, more complex tests of implicit and explicit coordination control of the mobile manipulator.