Guest Editorial Introduction To The Special Issue On Automatic Target Detection And Recognition

AUTOMATIC target recognition (ATR) generally refers to the autonomous or aided target detection and recognition by computer processing of data from a variety of sensors such as forward looking infrared (FLIR), synthetic aperture radar (SAR), inverse synthetic aperture radar (ISAR), laser radar (LADAR), millimeter wave (MMW) radar, multispectral/hyperspectral sensors, low-light television (LLTV), video, etc. It is an extremely important capability for targeting and surveillance missions of defense weapon systems operating from a variety of platforms. ATR is aimed at reducing the workload for human operators (e.g., image analysts, pilots, tankers) who are tasked with a large scope of activities ranging from assessing the battlefield/battlespace situation over large areas and volumes to targeting individual targets on land, sea, or air. ATR is also used to reacquire targets when used on unmanned lethal weapon systems such as missiles. Its need is dictated by large volumes of data requiring analysis and by the short timelines required by target acquisition scenarios. Its successful application will greatly increase the effectiveness and efficiency of weapon systems. ATR is a multidisciplinary field that requires diverse technology and expertise in sensors, processing algorithms, architectures, and evaluation of hardware and software systems. The major technical challenge for ATR is contending with the combinatorial explosion of target signature variations due to target configuration (e.g., stores, articulation, manufacturing, wear/tear), target/sensor acquisition parameters (e.g., aspect, depression, squint angles), target phenomenology (e.g., cavity responses, glints, infrared thermal behavior), and target/clutter interaction (e.g., foliage masking, camouflage). ATR systems must maintain low false alarm rates in the face of varying and complex backgrounds, and must operate in real time. Another extremely important challenge for ATR is the evaluation and prediction of ATR field performance given the practical limitation of data sets that cannot represent the extreme variability of the real world. The ability to rapidly insert new targets and to train algorithms on-the-fly in the field are important challenges to support flexible and sustained employment of ATR. The ATR field has evolved from using statistical pattern recognition approaches to model-based vision and to knowledge-based systems. Currently, adaptive and learning systems, focused on parts of the ATR problem, are also being developed. Current research areas include knowledge-based and model-based techniques, and approaches based on using physical principles, detection theory, multiresolution processing, statistical techniques, neural nets, and genetic algorithms. However, no single approach is likely to be the solution to all