Real-time Architecture of Stereo Vision for Robot Eye

Abstract

We present a vision system of high speed rectification and real time stereo matching fit for robots. It has a small hardware architecture so that it can be embedded in a large system at a low cost. It uses the global matching method and the parallel architecture to perform high quality and real time stereo matching. This system is based on a trellis form with a parallel optical axis and composed of pre and main processing modules. We implement the stereo chip on Xilinx FPGA with 60 processing elements that can obtain disparity range of 60 levels