Multisensor Fusion using FLIR and LADAR Identification

摘要

The primary objective of this project was to provide the Surrogate SemiautonomousVehicle (SSV) with a Demonstration II capability of performing automatic target recognition/identification (ATR/I). Detected objects of interest would be imaged with FLIR and/or LADAR sensors so the ATR/I algorithms must be compatible with either sensor, as well as exploit the synergy of processing both sensors simultaneously. Our approach does not rely upon the precise coregistration of multiple sensors, but rather performs geometric hashing on the individual FLIR and LADAR images. Hashing represents an object by a collection of points, which are then matched to similarly constructed models. The matching is accomplished by iteratively selecting pairs of points, placing them in a Euclidean geometry coordinate system, concurrently translating and rotating all other object points to the same geometry, and then counting the number of occurrences of object and model points in the same cell. The geometric hashing software, originally developed for 2D SAR and FLIR imagery has been extended to also accommodate 3D Ladar range and intensity imagery. The 2D hashing software was modified to allow up to ten dimensions. Currently, a 4D scheme is being used which represents (x,y) position, range, and intensity features.