The Design Of Stereo Vision Real-time Image Matching System Based On FPGA

With the rapid development of information science, the technology of stereo vision image processing has a considerable progress. At present, the stereo vision image processing technology has been widely used in aerospace, automotive electronics, medical imaging and industrial production, etc. Digital image processing features below: processing of large amount of data, the handling is very time-consuming, the system real time not easy to meet. As one of the key technology of stereo vision image matching (stereo matching), can effectively solve the problem seriously influenced the development of stereo vision. This paper studies the FPGA to use hardware description language realized the binocular stereo image matching, through the function module of the hardware, solve the problem of the video image processing speed. With the rapid development of microelectronics technology, more excellent FPGA performance, unceasingly rich resources, is stereo image signal processing in the algorithm, the system architecture brought more superior in the solution.This paper comes from XX company cooperation project:double camera image depth mapping. That is based on FPGA binocular stereo vision of the real-time stereoscopic imaging system design and implementation, this thesis research focused on double camera stereo matching function module algorithm analysis and selecting and the optimization, matching module structure, implementing and simulation results show the scheme formulated, hardware resources occupation/performance analysis and module, the realization of the function of function/timing simulation, test and the writing of the module driver, systems integration, etc. The essence of the region algorithm is to use local window of information between gray relevant degree, performance relies on the choice of similarity measure, now commonly used to measure standards have matching area of gray poor summation of absolute value of pixels, gray sent square summation and matching piece of pixels in the normalized difference value is peace, etc.; And these algorithm complexity are very high, to deal with a large number of data, using the general computer or digital processing chip serial processing methods can not meet the rapid, real-time processing performance requirements, and so we need to develop a specialized hardware parallel processing system.The whole design which is achieved in Quartus Ⅱ10.1and third party MODELSIM simulation software development environment,is realized in the hardware of Altera DE2-115platform. The analysis results show that the use of FPGA hardware realize image matching can not only gain good treatment effect, but also satisfy the requirement of real-time.