Model-Based 3D LADAR ATR Development and Evaluation

"We propose to characterize performance levels of a model-based 3D LADAR ATR system in order to understand the applicability of LADAR sensors to target exploitation in complex scenes. The MSTAR program has successfully demonstrated the applicability ofmodel-based recognition to 2D SAR processing of targets in extended operating conditions (EOCs). We will thus extend that approach to 3D LADAR exploitation. 3D LADAR, by virtue of generating high resolution 3D geometric reconstructions of a scene, is anatural application for shape-oriented model-based reasoning. Our technical approach to 3D LADAR ATR consists of supporting single to multiple LADAR looks and is based on uncertainty modeling and hierarchical part-based surface alignment methods. The 3Drecognition approach is well-suited for recognizing partially-occluded variably-configured targets in cluttered environments and is therefore an excellent candidate for ATR evaluation under EOCs. Our ATR evaluation methodology will also leverage theexperience and tools developed under the MSTAR program, decomposing performance measures along the sensor, target, and environment operating dimensions. Our team is composed of the developers of MSTAR's model-based reasoning modules and the evaluators ofthe MSTAR system and is well-positioned to extend our SAR experience to LADAR performance characterization.The development of a structured evaluation process for 3D recognition technology has