Optimizing Aided Target-Recognition Performance

Automatic target recognition (ATR) is a challenging task for several computer vision applications. It requires efficient, accurate, and robust methods for target detection and target identification. Deep learning has shown great success in many computer vision applications involving color RGB images. However, the performance of these networks in ATR with infrared sensor data needs further investigation. In this paper, we propose a multistage automatic target detection and recognition (ATDR) system that performs both target detection and target classification on infrared (IR) imagery using deep learning. Our system processes large IR image frames where targets take <1 % of the total number of pixels. First, we train a state-of-the-art object detector you only look once (YOLO) to localize all potential targets in the input image frame. Then, we train a convolutional neural network (CNN) to identify these detections as targets or false alarms. In this second phase, we adapt and analyze the performance of three CNN architectures: a compact and fully connected CNN, VGG16 with batch normalization, and a wide residual neural network (WRN). We also explore the use of a loss function that optimizes directly the area under the receiver operating characteristic (ROC) curve (AUC), and adapt it to our ATR application. To enhance the robustness of the proposed ATR to perturbation and variations introduced during the detection stage, we train our CNN classifiers on automatically detected targets using YOLO, in addition to ground truth bounding boxes and apply selected data augmentation techniques. To simulate real testing environments, where the spatial location of the targets within the image frame is unknown, only YOLO-detected boxes are used during validation. We evaluate our ATDR on a real benchmark dataset that includes different vehicles captured at different resolutions. Our experiments have shown that YOLO can detect most of the targets at the expense of generating a high number of false alarms. We show that the VGG-16 network with batch normalization, which is the best performing model, can correctly identify the classes of the targets, as well as classify the majority of YOLO’s false detections into an additional nontarget class. We also show that the proposed training modification to optimize an AUC-based loss function for ATR proved to be advantageous mainly in identifying difficult targets.

Performance of automatic target recognition (ATR) systems depends on numerous factors including the mission description, operating conditions, sensor modality, and ATR algorithm itself. Performance prediction models sensitive to these factors could be applied to ATR algorithm design, mission planning, sensor resource management, and data collection design for algorithm verification. Ideally, such a model would return measures of performance (MOPs) such as probability of detection (Pd), correct classification (Pc), and false alarm (Pfa), all as a function of the relevant predictor variables. Here we discuss the challenges of model-based and data-based approaches to performance prediction, concentrating especially on the synthetic aperture radar (SAR) modality. Our principal conclusion for model-based performance models (predictive models derived from fundamental physics- and statistics-based considerations) is that analytical progress can be made for performance of ATR system components, but that performance prediction for an entire ATR system under realistic conditions will likely require the combined use of Monte Carlo simulations, analytical development, and careful comparison to MOPs from real experiments. The latter are valuable for their high-fidelity, but have a limited range of applicability. Our principal conclusion for data-based performance models (that fit empirically derived MOPs) offer a potentially important means for extending the utility of empirical results. However, great care must be taken in their construction due to the necessarily sparse sampling of operating conditions, the high-dimensionality of the input space, and the diverse character of the predictor variables. Also the applicability of such models for extrapolation is an open question.

Automatic target recognition (ATR) of synthetic aperture radar (SAR) images is performed on either global or local features. The global features can be extracted and classified with high efficiency. However, they lack the reasoning capability thus can hardly work well under the extended operation conditions (EOCs). The local features are often more difficult to extract and classify but they provide reasoning capability for EOC target recognition. To combine the efficiency and robustness in an ATR system, a hierarchical fusion of the global and local features is proposed for SAR ATR in this paper. As the global features, the random projection features can be efficiently extracted and effectively classified by sparse representation-based classification (SRC). The physically relevant local descriptors, i.e., attributed scattering centers (ASCs), are employed for local reasoning to handle various EOCs like noise corruption, resolution variance, and partial occlusion. A one-to-one correspondence between the test and template ASC sets is built by the Hungarian algorithm. Then, the local reasoning is performed by evaluating individual matched pairs as well as the false alarms and missing alarms. For the test image to be recognized, it is first classified by the global classifier, i.e., SRC. Once a reliable decision is made, the whole recognition process terminates. When the decision is not reliable enough, it is passed to the local classifier, where a further classification by ASC matching is carried out. Therefore, by the hierarchical fusion strategy, the efficiency of global features and the robustness of local descriptors to various EOCs can be maintained jointly in the ATR system. Extensive experiments on the moving and stationary target acquisition and recognition data set demonstrate that the proposed method achieves superior effectiveness and robustness under both SOC and typical EOCs, i.e., noise corruption, resolution variance, and partial occlusion, compared with some other SAR ATR methods.

This paper deals with several original contributions to an automatic target recognition (ATR) system, which is applied to underwater mine classification. The contributions concentrate on feature selection and object classification. First, a sophisticated filter method is designed for the feature selection. This filter method utilizes a novel feature relevance measure, the composite relevance measure (CRM). Feature relevance measures in the literature (e.g., mutual information and relief weight) evaluate the features only with respect to certain aspects. The CRM is a combination of several measures so that it is able to provide a more comprehensive assessment of the features. Both linear and nonlinear combinations of these measures are taken into account. A wide range of classifiers is able to provide satisfactory classification results by using the features selected according to the CRM. Second, in the step of object classification, an ensemble learning scheme in the framework of the Dempster-Shafer theory is introduced to fuse the results obtained by different classifiers. This fusion can improve the classification performance. We propose a reasonable construction of the basic belief assignment (BBA). The BBA considers both the reliability of the classifiers and the support of individual classifiers provided to the hypotheses about the types of test objects. Finally, this ATR system is applied to real synthetic aperture sonar imagery to evaluate its performance.

Proper evaluation of a pattern recognition system in the lab is paramount to its success in the field. In most commercial pattern recognition applications, such as breast cancer detection, optical character recognition, and industrial quality assurance, the boundaries and expectation of the system are well defined. This is due, at least in part, to an excellent understanding of the problem and data space for these applications. For these functions, a method for rigorous evaluation is well understood. However, the size and complexity of the data and problem spaces for automatic target recognition (ATR) systems is enormous. The consequences are that a complete understanding of how an ATR system will perform in practice is extraordinarily difficult to estimate. Thus the act of evaluating an ATR system becomes as important as its design. This paper compiles and reports the techniques used to evaluate ATR system performance. It surveys the specific difficulties associated with ATR performance estimation as well as approaches used to mitigate these obstacles.

The Deputy Under Secretary of Defense for Science and Technology (DUSD/S&T), as part of their ongoing ATR Program, has sponsored an effort to develop and demonstrate methods for evaluating ATR algorithms that utilize multiple data sources, i.e., fusion-based ATR. The AFRL COMPASE Center has formed a strong ATR evaluation team and this paper presents results from this program, focusing on the human-in-the-loop, i.e. assisted image exploitation. Reliance on Automated Target Recognition (ATR) technology is essential to the future success of Intelligence, Surveillance, and Reconnaissance (ISR) missions. Often, ATR technology is designed to aid the analyst, but the final decision rests with the human. Traditionally, evaluation of ATR systems has focused mainly on the performance of the algorithm. Assessing the benefits of ATR assistance for the user raises interesting methodological challenges. We will review the critical issues associated with evaluations of human-in-the-loop ATR systems and present a methodology for conducting these evaluations. Experimental design issues addressed in this discussion include training, learning effects, and human factors issues. The evaluation process becomes increasingly complex when data fusion is introduced. Even in the absence of ATR assistance, the simultaneous exploitation of multiple frames of co-registered imagery is not well understood. We will explore how the methodology developed for exploitation of a single source of data can be extended to the fusion setting.

For the Automatic Target Recognition (ATR) algorithm, the quality of the input image sequence can be a major determining factor as to the ATR algorithm's ability to recognize an object. Based on quality, an image can be easy to recognize, barely recognizable or even mangled beyond recognition. If a determination of the image quality can be made prior to entering the ATR algorithm, then a confidence factor can be applied to the probability of recognition. This confidence factor can be used to rate sensors; to improve quality through selectively preprocessing image sequences prior to applying ATR; or to limit the problem space by determining which image sequences need not be processed by the ATR algorithm. It could even determine when human intervention is needed. To get a flavor for the scope of the image quality problem, this paper reviews analog and digital forms of image degradation. It looks at traditional quality metric approaches such as peak signal-to-noise ratio. It examines a newer metric based on human vision data, a metric introduced by the Institute for Telecommunication Sciences. These objective quality metrics can be used as confidence factors primarily in ATR systems that use image sequences degraded due to transmission systems. However, to determine the quality metric, a transmission system needs the original input image sequence and the degraded output image sequence. This paper suggests a more general approach to determining quality using analysis of spatial and temporal vectors where the original input sequence is not explicitly given. This novel approach would be useful where there is no transmission system but where the ATR system is part of the sensor, on-board a mobile platform. The results of this work are demonstrated on a few standard image sequences.

: This report focuses on the development of an automatic target recognition (ATR) system using high resolution synthetic aperture radar (SAR) imagery. The system achieves 95 to 100 percent recognition rates when applied to a set of MSTAR images. Typically, the system takes less than one minute to match an input image to a candidate vehicle class with Matlab programs running on a Pentium II 300 MHz machine. Experiments based on conventional recognition techniques were conducted for comparisons. Study of persistent scattering confirms the feasibility of implementing a SAR ATR system using physical image features. A new generic vehicle model, parameterized by the length, width, and orientation of a target is used in a two-phase recognition process with hypothesis generation and verification aimed at addressing the combinatorial target recognition problem. In the hypothesis generation stage, a few likely candidate target classes are identified from a target database with positive evidence. The candidates are assessed using both positive and negative evidence in the hypothesis verification stage. Leading surface estimation, image alignment, Delaunay walk, and recognition metrics are introduced to improve performance of the SAR ATR system.

We present a performance analysis of an indexing module contained in an automatic target recognition (ATR) system developed at Worcester Polytechnic Institute (WPI). The linear signal decomposition/direction of arrival (LSD/DOA) technique provides a low computational-cost indexing function for pose in ATR applications. The LSD/DOA technique forms a reduced-size, essential-information object data set which implicitly incorporates target and sensor variation specific data. To control the computational costs, the system consists of two independent components: a computationally expensive off-line component which forms the object representation and a computationally inexpensive on-line object recognition component. The size of the stored data set may also be adjusted providing a means to trade off complexity versus performance. Synthetic aperture radar (SAR) data collected as part of the Moving and Stationary Target Acquisition and Recognition (MSTAR) program has been released to the public, providing an opportunity for ATR performance assessment against a standard high quality data set. We present the results of a number of simulated tests against the MSTAR public database to demonstrate the performance of the LSD/DOA index module on a given target along with the performance of the WPI ATR itself as a function of the depression angle and target configuration.

Forward Looking Infrared (FLIR) automatic target recognition (ATR) systems depend upon the capacity of the atmosphere to propagate thermal radiation over long distances. To date, not much research has been conducted on analyzing and mitigating the effects of the atmosphere on FLIR ATR performance, even though the atmosphere is often the limiting factor in long-range target detection and recognition. The atmosphere can also cause frame-to-frame inconsistencies in the scene, affecting the ability to detect and track moving targets. When image quality is limited by turbulence, increasing the aperture size or improving the focal plane array cannot improve ATR performance. Traditional single frame image enhancement does not solve the problem. A new approach is described for reducing the effects of turbulence. It is implemented under a lucky-region-imaging framework using short integration time and spatial domain processing. It is designed to preserve important target and scene structure. Unlike previous Fourier-based approaches originating from the astronomical community, this new approach is intended for real-time processing from a moving platform, with ground as the background. The system produces a video stream with minimal delay. A new video quality measure (VQM_turb) is presented for quantifying the success of turbulence mitigation on real data where no reference imagery is available. The VQM_turb is the core of the innovation because it allows a wide range of algorithms to be quantitatively compared. An algorithm can be chosen, and then tuned, to best-fit available processing power, latency requirements, scenarios and sensor characteristics.

A research platform with four cameras in the infrared and visible spectral domains is under development at the Norwegian Defence Research Establishment (FFI). The platform will be mounted on a high-speed jet aircraft and will primarily be used for image acquisition and for development and test of automatic target recognition (ATR) algorithms. The sensors on board produce large amounts of data, the algorithms can be computationally intensive and the data processing is complex. This puts great demands on the system architecture; it has to run in real-time and at the same time be suitable for algorithm development. In this paper we present an architecture for ATR systems that is designed to be exible, generic and efficient. The architecture is module based so that certain parts, e.g. specific ATR algorithms, can be exchanged without affecting the rest of the system. The modules are generic and can be used in various ATR system configurations. A software framework in C++ that handles large data ows in non-linear pipelines is used for implementation. The framework exploits several levels of parallelism and lets the hardware processing capacity be fully utilised. The ATR system is under development and has reached a first level that can be used for segmentation algorithm development and testing. The implemented system consists of several modules, and although their content is still limited, the segmentation module includes two different segmentation algorithms that can be easily exchanged. We demonstrate the system by applying the two segmentation algorithms to infrared images from sea trial recordings.

This paper is concerned with the development of a framework in which reasonable bounds and approximations for the performance of automatic target recognition (ATR) systems may be obtained. The relative merits of evaluations focusing on various components of ATR systems are discussed. Techniques in which structural knowledge of the radar target (embodied in phenomenological and pattern models) might be used to characterize representations in pattern space, and distances in decision space are described. Preliminary results corresponding to various assumptions regarding structural constraints are presented.

The resolution capability of imaging systems is effected by a blurring effect due to the vibration and motion recorded in the image. This disturbance is often more severe than electronic and optical limitations inherent in the system. This fact must be considered when dealing with the development and analysis of automatic target recognition (ATR) systems for military applications. The aim of this research is to analyze the influence of image vibrations and motion upon the probability of acquiring the target with an ATR system. The analysis includes factors that characterize the relationship existing between the target and its background. A high level of correlation is expected between these factors and the probability of target detection, enabling efficient performance in the prediction and evaluation of any ATR system. The results of this research can be implemented in military applications as well as in developing image restoration procedures for image-blur conditions.

Low altitude, small radar cross-section (RCS), and slow speed (LSS) targets, for example small unmanned aerial vehicles (UAVs), have become increasingly significant. In this paper, we propose a new automatic target recognition (ATR) system and a complete ATR chain based on multi-dimensional features and multi-layer classifier system using L-band holographic staring radar. We consider all steps of the processing required to make a classification decision out of the raw radar data, mainly including preprocessing for the raw measured Doppler data including regularization and main frequency alignment, selection, and extraction of effective features in three dimensions of RCS, micro-Doppler, and motion, and multi-layer classifier system design. We design creatively a multi-layer classifier system based on directed acyclic graph. Helicopters, small fixed-wing, and rotary-wing UAVs, as well as birds are considered for classification, and the measured data collected by L-band radar demonstrates the effectiveness of the proposed complete ATR classification system. The results show that the ATR classification system based on multi-dimensional features and k-nearest neighbors (KNN) classifier is the best, compared with support vector machine (SVM) and back propagation (BP) neural networks, providing the capability of correct classification with a probability of around 97.62%.

Automatic target recognition (ATR) is an important part for many computer vision applications. Despite the extensive research which has been carried out in this area for many years, there is no ATR system which performs well on all applications. Recently, different object recognition frameworks have been proposed which yield a high performance in baseline databases. However, our experiments showed that they can fail in real-world scenarios, when dealing with a limited number of data samples. In this paper, we propose a new ATR system, based on deep convolutional neural network (DCNN), to detect the targets in forward looking infrared (FLIR) scenes and recognize their classes. In our proposed ATR framework, a fully convolutional network is trained to map the input FLIR imagery data to a fixed stride correspondingly-sized target score map. The potential targets are identified by applying a threshold on the target score map. Finally, the corresponding regions centered at these target points are fed to a DCNN to classify them into different target types while at the same time rejecting the false alarms. The proposed architecture achieves a significantly better performance in comparison with that of the state-of-the-art methods on two large FLIR image databases.