Model-Based Automatic Target Recognition System for the UGV/RSTA Ladar

摘要

This paper describes our initial work under the ARPA Unmanned Ground Vehicle (UGV) Demo II Program in the reconnaissance, surveillance, and target acquisition (RSTA) area. We are developing the automatic target recognition (ATR) system that will process the imagery from the RSTA laser radar (ladar). A real-time demonstration of this capability is scheduled for the "Demo Ⅱ" of 1996. We discuss all major components of this end-to-end ATR system and provide more details about critical elements we have built and tested so far. A major topic of interest is the use of sets of azimuth-dependent functional templates and functional template correlation for detecting and recognizing targets in vertically-reprojected forward-looking ladar data. Other topics are the use of height-limited verticality as an interest image for focusing attention and of the Hough transform for getting a preliminary estimate of target orientation. Since the RSTA ladar remains to be procured and built, we also describe alternate sources of data we have used to develop and test the system elements built to date.