We have developed a wide variety of ATR algorithms for different sensing modalities including: Forward Looking Infra-Red (FLIR), Synthetic Aperture Radar (SAR), airborne Laser Radar (LADAR), Ground Penetrating Radar (GPR), video surveillance, seismic, acoustic, and others.

These have been used to detect and identify a wide variety of targets including: tanks, automobiles, missile launchers, landmines, etc. The basic approach used in many of these applications involves three steps: (1) non-linear image pre-processing for noise reduction; (2) development of multiple target detection algorithms; and (3) non-linear fusion of multi-detector outputs for optimal detection and recognition.

An extremely robust ATR algorithm developed in our research is the Morphological Shared Weight Neural Networks (MSNN). The MSNN is a multi-layer neural network that combines both feature extraction (based on shape) and classification/identification capabilities. Thus it can simultaneously be used for both target detection and recognition.

Examples of the MSNN ATR algorithm applied to airborne LADAR, SAR, and video surveillance imagery are shown below.