Abstract: ■ Although a number of object-recognition techniques have been developed to process terrain scenes scanned by laser radar (ladar), these techniques have had limited success in target discrimination, in part due to low-resolution data and limits in available computation power. We present a pose-independent automatic target detection and recognition system that uses data from an airborne threedimensional imaging ladar sensor. The automatic target recognition system uses geometric shape and size signatures from target models to detect and recognize targets under heavy canopy and camouflage cover in extended terrain scenes. The system performance was demonstrated on five measured scenes with targets both out in the open and under heavy canopy cover, where the target occupied between 1% to 10% of the scene by volume. The automatic target recognition section of the system was successfully demonstrated for twelve measured data scenes with targets both out in the open and under heavy canopy and camouflage cover. Correct target identification was also demonstrated for targets with multiple movable parts in arbitrary orientations. The system achieved a high recognition rate along with a low false-alarm rate. Immediate benefits of the presented work are in the area of automatic target recognition of military ground vehicles, in which the vehicles of interest may include articulated components with variable position relative to the body, and may come in many possible configurations. Other application areas include human detection and recognition for homeland security, and registration of large or extended terrain scenes.

Abstract: A synthetic aperture radar (SAR) automatic target recognition (ATR) system based on the minimum noise and correlation energy (MINACE) distortion-invariant filter (DIF) is presented. A set of MINACE filters covering different aspect ranges is synthesized for each object using a training set of images of that object and a validation set of confuser and clutter images. No prior DIF work addressed confuser rejection. We also address use of fewer DIFs per object than prior work did. The selection of the MINACE filter parameter c for each filter is automated using training and validation sets. The system is evaluated using images from the Moving and Stationary Target Acquisition and Recognition (MSTAR) public database. The classification scores (PC) and the number of false alarm scores for confusers and clutter (PFA and PCFA respectively) are presented for the benchmark three-class MSTAR database with object variants and two confusers. The pose of the input test image is not assumed to be known, thus the problem addressed is more realistic than in prior work, since pose estimation of SAR objects has a large margin of error. Results for both confuser and clutter rejection are presented.

Abstract: This paper describes an algorithm for the detection of low- and high-contrast targets in forward-looking infrared imagery while rejecting the effects of clutter and other associated detrimental factors. The proposed automatic target recognition algorithm involves two modules—a detector module and a clutter rejection module. The detection algorithm, based on morphology-based preprocessing, acts as a prescreener that selects possible candidate target regions for further analysis and places target-size markers in those preselected regions. The application of simple nonlinear grayscale operations in the proposed detection algorithm has been found to be especially suitable for real-time implementation. The clutter rejection module uses target- and background-specific information, extracted from training sample, to reduce false alarms often generated in the detection step. The application of two Mahalonobis distances, derived from target and background features of the training image, improves false-alarm rejection. Preliminary results indicate that the developed detection and clutter rejection modules exhibit excellent detection performance for both low- and high-contrast targets in complex backgrounds while ensuring a low false-alarm rate.

Abstract: A hidden Markov model (HMM)-based method for recognizing aerial targets according to the sequential high-range-resolution (HRR) radar signature is presented. Its recognition features are the location information of scattering centers extracted from the HRR radar echoes by the relax algorithm. The HMM is used to characterize the spatio-temporal information of a target. Several HMMs are cascaded in a chain to model the variation in the target orientation and used as classifiers. Computer simulations with the inverse synthetic aperture radar (ISAR) data are given to demonstrate that for an open-set recognition, average class-recognition rates of 84.50% and 89.88% are achieved, respectively, under two given conditions.

Abstract: Automatic target recognition from high range resolution radar profiles remains an important and challenging problem. In this paper, we present a novel feature set for this task that combines a noise-robust superresolution characterisation of the target scattering centres derived using the MUSIC algorithm with a representation of the target's radar shadow shape. To obtain the shadow shape features, three alternative spectral estimation methods are investigated. Using a hidden Markov model to represent aspect dependence, we demonstrate that the inclusion of the shadow features results in a significant improvement in recognition performance. Using azimuth apertures of 3/spl deg/ and 6/spl deg/ in a 10-target classification task from the MSTAR database, we obtain overall classification error rates of 1.3% and 0.2% respectively. These results are significantly better than those obtained by other published methods on the same database.

Abstract: A binary hierarchical classifier is proposed for automatic target recognition. We also require rejection of non-object (non-target) inputs, which are not seen during training or validation, thus producing a very difficult problem. The SVRDM (support vector representation and discrimination machine) classifier is used at each node in the hierarchy, since it offers good generalization and rejection ability. Using this hierarchical SVRDM classifier with magnitude Fourier transform (|FT|) features, which provide shift-invariance, initial test results on infrared (IR) data are excellent.

Abstract: A target detection method from low contrast forward looking infrared (FLIR) images is proposed. It is known that detecting small targets in remotely sensed image is difficult and challenging work. The goal is to identify target areas with small number of false alarms in a thermal infrared scene of battlefield. The proposed method consists of following three stages. First, center-surround difference with local adaptive threshold is proposed in order to find salient areas in an input image. Second, local thresholding is proposed to the local region of interest (ROf) based on the result of first step. The second step is needed to segment target silhouettes precisely. Third, the extracted binary target silhouettes are compared with target template using size and affinity to remove clutters. In the experiments, many natural infrared images with high variability are used to prove performance the proposed method. It is compared with a morphological method using receiver operating characteristic (ROC) curve and execution time. The result shows that our method is superior to the morphological method and it can be applied to automatic target recognition (ATR) system.

Abstract: This paper presents a new automatic target detection (ATD) algorithm to detect targets such as battle tanks and armored personal carriers in ground-to-ground scenarios. Whereas most ATD algorithms were developed for forward-looking infrared (FLIR) images, we have developed an ATD algorithm for charge-coupled device (CCD) images, which have superior quality to FLIR images in daylight. The proposed algorithm uses fast binary template matching with an adaptive binarization, which is robust to various light conditions in CCD images and saves computation time. Experimental results show that the proposed method has good detection performance.

Abstract: We propose a novel automatic target recognition (ATR) system for classification of three types of ground vehicles in the MSTAR public release database. First, each image chip is pre-processed by extracting fine and raw feature sets, where raw features compensate for the target pose estimation error that corrupts fine image features. Then, the chips are classified by using the adaptive boosting (AdaBoost) algorithm with the radial basis function (RBF) net as the base learner. Since the RBF net is a binary classifier, we decompose our multiclass problem into a set of binary ones through the error-correcting output codes (ECOC) method, specifying a dictionary of code words for the set of three possible classes. AdaBoost combines the classification results of the RBF net for each binary problem into a code word, which is then "decoded" as one of the code words (i.e., ground-vehicle classes) in the specified dictionary. Along with classification, within the AdaBoost framework, we also conduct efficient fusion of the fine and raw image-feature vectors. The results of large-scale experiments demonstrate that our ATR scheme outperforms the state-of-the-art systems reported in the literature.

Abstract: The threat of hostile surveillance and weapon systems require military aircraft to fly under extreme conditions such as low altitude, high speed, poor visibility and incomplete terrain information. The probability of collision with natural and man-made obstacles during such contour missions is high if detection capability is restricted to conventional vision aids. Forward-looking scanning laser radars which are build by the EADS company and presently being flight tested and evaluated at German proving grounds, provide a possible solution, having a large field of view, high angular and range resolution, a high pulse repetition rate, and sufficient pulse energy to register returns from objects at distances of military relevance with a high hit-and-detect probability. The development of advanced 3d-scene analysis algorithms had increased the recognition probability and reduced the false alarm rate by using more readily recognizable objects such as terrain, poles, pylons, trees, etc. to generate a parametric description of the terrain surface as well as the class, position, orientation, size and shape of all objects in the scene. The sensor system and the implemented algorithms can be used for other applications such as terrain following, autonomous obstacle avoidance, and automatic target recognition. This paper describes different 3D-imaging ladar sensors with unique system architecture but different components matched for different military application. Emphasis is laid on an obstacle warning system with a high probability of detection of thin wires, the real time processing of the measured range image data, obstacle classification und visualization.

Abstract: Evidence theory is widely used in automatic target recognition(ATR) system.One of the problems in real application is that not only the observation collected by sensors,but also the attributes of targets in the model database may be fuzzy too.In this situation,how to automatically determine the mass function of fuzzy information is an open issue.In this paper,a method of automatically determining mass function for target recognition is presented.After representing both the individual attribute of target in the model database and the sensor observation or report as fuzzy membership function,a random sets model of this fuzzy information is introduced.Then,a likelihood function is obtained to deal with the fuzzy data collected by each sensor.The likelihood function has probabilistic character and can be transformed into a mass function,which numerically shows the support degree of the hypotheses that the target is a certain target under the collected fuzzy information.The present approach has been tested in an ATR system to illustrate its efficiency and can be easily used in many fuzzy information fusion systems.

Abstract: We propose a novel automatic target recognition (ATR) system for classification of three types of ground vehicles in the MSTAR public release database. First, MSTAR image chips are represented as fine and raw feature vectors, where raw features compensate for the target pose estimation error that corrupts fine image features. Then, the chips are classified by using the adaptive boosting (AdaBoost) algorithm with the radial basis function (RBF) network as the base learner. Since the RBF network is a binary classifier, we decompose our multiclass problem into a set of binary ones through the error-correcting output codes (ECOC) method, specifying a dictionary of code words for the set of three possible classes. AdaBoost combines the classification results of the RBF network for each binary problem into a code word, which is then "decoded" as one of the code words (i.e., ground-vehicle classes) in the specified dictionary. Along with classification, within the AdaBoost framework, we also conduct efficient fusion of the fine and raw image-feature vectors. The results of large-scale experiments demonstrate that our ATR scheme outperforms the state-of-the-art systems reported in the literature.

Abstract: An automatic target recognition (ATR) algorithm, based on greedy learning of Gaussian mixture model (GMM) is developed in this work. The GMMs were obtained for a wide range of ground surveillance radar targets such as: walking person(s), tracked or wheeled vehicles, animals and clutter. Maximum-likelihood (ML) and "majority voting" decision schemes were applied to these models for target classification. The corresponding classifiers were trained and tested using distinct databases of target echoes, recorded by ground surveillance radar. ML and "majority voting" classifiers obtained classification rates of 88% and 96%, correspondingly. Both classifiers outperform trained human operators.

Abstract: Automatic target recognition from high range resolution radar profiles remains an important and challenging problem. In this paper, we present a novel feature set for this task that combines a representation of the target's radar shadow with a noise-robust superresolution characterisation of the target scattering centres derived from the MUSIC algorithm. Using an HMM to represent aspect dependence, we demonstrate that the inclusion of the shadow features results in a significant improvement in recognition performance. We evaluate our proposed feature set on a closed-set identification task using targets from the MSTAR database and show that it results in lower recognition error rates than previously published methods using the same data.

Abstract: The effective recognition and tracking for extended target in the complex background is a challenged problem. This paper classifies the algorithm of current Automatic Target Recognition (ART). Those features, which have RST invariance and tracking method of moving target, are also discussed. At last, some issues that should be solved in the future are proposed.

Abstract: This paper presents our automated filter-synthesis algorithm for the minimum noise and correlation energy (MINACE) distortion-invariant filter (DIF). We discuss use of this autoMinace filter in face recognition and automatic target recognition (ATR), in which we consider both true-class object classification and rejection of non-database objects (impostors in face recognition and confusers in ATR). We use at least one Minace filter per object class to be recognized; a separate Minace filter or a set of Minace filters is synthesized for each object class. The Minace parameter c trades-off distortion-tolerance (recognition) versus discrimination (impostor/confuser/clutter rejection) performance. Our automated Minace filter-synthesis algorithm (autoMinace) automatically selects the Minace filter parameter and selects the training set images to be included in the filter, so that we achieve both good recognition and good impostor/confuser and clutter rejection performance; this is achieved using a training and validation set. No impostor/confuser, clutter or test set data is present in the training or validation sets. Use of the peak-to-correlation energy (PCE) ratio is found to perform better than the correlation peak height metric. The use of circular versus linear correlations is addressed; circular correlations require less storage and fewer online computations and are thus preferable. Representative test results for three different databases - visual face, IR ATR, and SAR ATR - are presented. We also discuss an efficient implementation of Minace filters for detection applications, where the filter template is much smaller than the input target scene.

Abstract: A primary strength of the XCS approach is its ability to create maximally accurate general rules. In automatic target recognition (ATR) there is a need for robust performance beyond so-called standard operating conditions (SOCs, those conditions for which training data is available) to extended operating conditions (EOCs, conditions of known targets that cannot be foreseen and trained for). EOCs include things like vehicle-specific variations, environmental effects (mud, etc.), unanticipated viewing angles, and articulation of components of the target (hatches, turrets, etc.). This paper presents experiments where XCS addresses structural generalization over global and local features normally used in ATR classification. In many SOCs, these features are adequate for target recognition. Our goal with XCS is to form generalized rules that utilize these features for effective ATR in EOCs. Results show that XCS is effective in this generalization task. Conclusions and future directions for research are discussed.

Abstract: Detecting target with low signal to noise ratio is an fundamental technique used for automatic target recognition (ATR) in infrared imagery,and its performances make an ultimate impact on detection sensitivity and effective distance of a systemIt is a leading key technique to indicate the ability of recognizing low observable target in infrared imageryAdaptive background estimation method is an efficient avenue to complete this taskOn the basis of summarizing several current estimation means,a novel morphological filtering algorithm improved properly is proposed in this paperSome theoretical analyses and experimental results show that this method is able to simplify operation of morphological conversion and to optimize formation of structuring elementsConsequently it can enhance the filtering result and accelerate the speed of operation as wellMoreover it is capable of preserving the property to protect signal characteristic and improving the inherent limitation not to be sensitive on fluctuation of noise and having better ability of adaptive background perception in morphological filtering algorithmIn conclusion this method is concise and efficientIt can provide good filtering results and robust adaptability to image targets with clutter background

Abstract: An automatic target recognition algorithm for synthetic aperture radar (SAR) imagery data is developed. The algorithm classifies an unknown target as one of the known reference targets based on a maximum likelihood estimation procedure. The algorithm helps assess and optimize the favorable effects of multiple image features on recognition accuracy. This study addresses four procedures: (1) feature extraction, (2) training set creation, (3) classification of unknown images, and (4) optimization of recognition accuracy. A three-feature probabilistic method based on extracted edges, corners, and peaks is used to classify the targets. Once the three features are extracted from the target image, binary images are created from each. Training sets, which are used to classify an unknown target, are then created using average Hausdorff distance values for each of the known members of the eight target image types (ZSU-23-4, ZIL131, D7, 2S1, SLICY, BDRM2, BTR60, and T62) included in the publicly available MSTAR test data. The average Hausdorff distance values are acquired from unknown target feature images and are compared to each training set. Each comparison provides the likelihood of the unknown target belonging to one of the eight possible known targets. For each target, eight likelihoods (for eight possible unknown targets) are determined based on the Hausdroff distances and the pre-assigned feature weights. The unknown target is then classified into the target type that has the maximum likelihood estimation value.

Abstract: Dr. David Cyganski is a Professor in the ECE Department at WPI where he performs research and teaches in the areas of linear and non-linear multidimensional signal processing, communications and computer networks, and supervises the WPI Convergent Technology Center. He is an active researcher in the areas of radar imaging, automatic target recognition, machine vision and protocols for computer networks. He is a coauthor of the book Information Technology: Inside and Outside. He has held the administrative positions of Vice President of Information Systems and Vice Provost at WPI.

Abstract: As the generalization of the classical Fourier transform, fractional Fourier transform(FRFT) is introduced into the field of speaker recognition in this paper. The individual feature sets derived from fractional Fourier transform achieve the excellent recognition success rate which goes up to the extent a little higher than the counterparts of the classical MFCC parameters when applied in the GMM classifiers. In addition, the computation efficiency of the feature extraction process arrives at the acceptable level which completely matches the one of MFCC parameter acquirement.

Abstract: Automatic target recognition (ATR) performance is a function of image quality and its representation in the signature model generation and used in the ATR training process. This paper reports ATR performance as a function of synthetic aperture radar (SAR) image quality parameters including clutter-to-noise ratio (CNR) and multiplicative noise ratio (MNR). Images with specified image quality values were produced by introducing controlled degradations to the MSTAR public release data. Two different families of ATR algorithms, the statistical model-based classifier of DeVore, et al., and optimal tradeoff synthetic discriminant function (OTSDF) are applied to those data. Target classification accuracy was measured as a function of CNR/MNR for both the training and test data, indicating sensitivity of performance to a priori knowledge of these particular image quality parameters. Confusion matrices are expanded to include target aspect bins, providing visibility into performance as a function of aspect angle.

Abstract: With research on autonomous underwater vehicles for minehunting beginning to focus on cooperative and adaptive behaviours, some effort is being spent on developing automatic target recognition (ATR) algorithms that are able to operate with high reliability under a wide range of scenarios, particularly in areas of high clutter density, and without human supervision. Because of the great diversity of pattern recognition methods and continuously improving sensor technology, there is an acute requirement for objective performance measures that are independent of any particular sensor, algorithm or target definitions. This paper approaches the ATR problem from the point of view of information theory in an attempt to place bounds on the performance of target classification algorithms that are based on the acoustic shadow of proud targets. Performance is bounded by analysing the simplest of shape classification tasks, that of differentiating between a circular and square shadow, thus allowing us to isolate system design criteria and assess their effect on the overall probability of classification. The information that can be used for target recognition in sidescan sonar imagery is examined and common information theory relationships are used to derive properties of the ATR problem. Some common bounds with analytical solutions are also derived.

Abstract: Rather than emitting pulses, passive radar systems rely on "illuminators of opportunity," such as TV and FM radio, to illuminate potential targets. These systems are attractive since they allow receivers to operate without emitting energy, rendering them covert. Until recently, most of the research regarding passive radar has focused on detecting and tracking targets. This dissertation focuses on extending the capabilities of passive radar systems to include automatic target recognition. The target recognition algorithm described in this dissertation uses the radar cross section (RCS) of potential targets, collected over a short period of time, as the key information for target recognition. To make the simulated RCS as accurate as possible, the received signal model accounts for aircraft position and orientation, propagation losses, and antenna gain patterns. An extended Kalman filter (EKF) estimates the target's orientation (and uncertainty in the estimate) from velocity measurements obtained from the passive radar tracker. Coupling the aircraft orientation and state with the known antenna locations permits computation of the incident and observed azimuth and elevation angles. The Fast Illinois Solver Code (FISC) simulates the RCS of potential target classes as a function of these angles. Thus, the approximated incident and observed angles allow the appropriate RCS to be extracted from a database of FISC results. Using this process, the RCS of each aircraft in the target class is simulated as though each is executing the same maneuver as the target detected by the system. Two additional scaling processes are required to transform the RCS into a power profile (magnitude only) simulating the signal in the receiver. First, the RCS is scaled by the Advanced Refractive Effects Prediction System (AREPS) code to account for propagation losses that occur as functions of altitude and range. Then, the Numerical Electromagnetic Code (NEC2) computes the antenna gain pattern, further scaling the RCS. A Rician likelihood model compares the scaled RCS of the illuminated aircraft with those of the potential targets. To improve the robustness of the result, the algorithm jointly optimizes over feasible orientation profiles and target types via dynamic programming.

Abstract: Over the last two decades, researchers investigated various approaches for detection and classification of targets in forward looking infrared (FLIR) imagery using correlation based techniques. In this paper, a novel technique is proposed to recognize and track single as well as multiple identical and/or dissimilar targets in real life FLIR sequences using a combination of extended maximum average correlation height (EMACH) and polynomial distance classifier correlation filter (PDCCF). The EMACH filters are used for the detection stage and PDCCF filter is used for the classification stage for improving the detection and discrimination capability. The EMACH and PDCCF filters are trained a priori using target images with expected size and orientation variations. In the first step, the input scene is correlated with all the detection filters (one for each desired or expected target class) and the resulting correlation outputs are combined. The regions of interest (ROI) are selected from the input scene based on the regions with higher correlation peak values in the combined correlation output. In the second step, PDCCF filter is applied to these ROIs to identify target types and reject clutters/backgrounds based on a distance measure and a threshold. Moving target detection and tracking is accomplished by applying this technique independently to all incoming image frames. Independent tracking of target(s) from one frame to the other allows the system to handle complicated situations such as a target disappearing in few frames and then reappearing in later frames. This method has been found to yield robust performance for challenging FLIR imagery in terms of faster and accurate detection and classification as well as tracking of the targets.

Abstract: This paper describes ATR/I algorithms that were developed by Thales Optronique for the real-time automatic target recognition and identification of ground vehicles in a Air-to-Ground non cooperative context. The main principles of the algorithm based on an exhaustive comparison between the input image and the elements of a 〈 Model Data Set 〉 are: • To avoid the variability on the gray-levels of the target, the comparison is not performed directly on the input gray-level image but on an edge image which is obtained with a segmentation algorithm derived from the classical Canny-Deriche edge detector . • The selected architecture is chosen to repeat many times a single instruction rather than to execute only one time a lot of different instructions. Therefore, the comparisons between the input image and the elements of the 〈 Model Data Set 〉 are performed in 2D with a correlative technique. • The computation time is achieved thanks to a coarse to fine analysis with different levels of comparison : in a first stage, a simple comparison measure is used which enables quick selection of a preliminary list of potential hypotheses. This measure is discriminating enough to select a small number of hypotheses and robust enough to select the true hypothesis associated with the target. These selected hypotheses are then analyzed during a second stage of processing using a more refined measure, and thus more time consuming than the previous one, but which is applied on a significantly reduced number of hypotheses.

Abstract: Radar automatic target recognition (ATR) using high range resolution profiles (HRRPs) is addressed. A subband fusion structure is proposed based on wavelet packet transforming. Multiple adaptive Gaussian classifiers (AGCs) are built for each subband, and the outputs of each of the subband classifiers are combined to make a final decision. Compared with the traditional wideband recognition approach, i.e., single band approach, the proposed approach can achieve better recognition performance, and is more robust to noise as well. Example results, based on the measured data, show the efficiency of the proposed method.