Real-time 3D Feature Extraction Hardware Algorithm with Feature Point Matching Capability

This paper proposes a real-time 3D feature extraction hardware algorithm with feature point matching capability between neighboring frames, which realizes 3D tracking of moving objects. This hardware algorithm is based on a 3D voting method. Both the 3D voting and tlie featiire point matching arc directly carried out through highly parallel processing by content addressable memory (CAM) i11 real time. For the 3D voting, the CAM acts as a PE (Processing Element) array that performs highly parallel processing. For tlie feature point matching, the CAM executes a highly parallel processing of nearest neighbor search. Sinliilations of CAM hardware size, processing time and accuracy show that real-time 3D feature extraction and feature point matching can be achieved using a single CAM chip with current VLSI technology. This CAM-bwed algorithm promises to be an important step towards the realization of a real-time and compact 3D tmcking system for moving objects.