Detecting mobile land targets in LADAR imagery with fuzzy algorithms

Abstract

We describe an automatic target detection system whose objective is to detect surface-to-air missile launchers and tanks -- small, mobile land targets. The system uses range and intensity images from a Laser Detection and Ranging (LADAR) sensor. Each type of image is filtered and passed to a pair of (different) detectors. Outputs from the four detectors are subsequently filtered with a Choquet fuzzy integral, and then fused with a generalized mean. Finally, blob detections are superposed on intensity images for visual assessment, and blob windows are compared to ground truth to compute performance statistics. The system is illustrated using images from two scenes available in several volumes of the Joint Advanced Weapons Guidance data; a target board site, and the so-called "Chicken Little" outer compound. The main objective of this article is to document the proposed architecture -- a novel target detection approach based on fuzzy decision theory -- as a proof of concept. We have left detailed comparisons with other approaches, as well as concrete error rate analysis, to another investigation.