基于竞争机制的双目视觉匹配与实时测距.pdfVIP

基于竞争机制的双目视觉匹配与实时测距 熊超 田小芳陆起涌 (复旦大学电子工程系，上海200433) 摘要利用立体视觉系统实现移动机器人避障具有探测范围宽，目标信息完整等优势。从立体视觉中提取距离信息是 计算机视觉的热点问题，其困难在于图像匹配问题。文章提出了一种基于竞争机制实现立体视觉匹配的方法。该方法将 匹配问题映射为一个矩阵。其元素代表可能匹配的像点。根据匹配约束，匹配矩阵元素之间竞争，最后胜出者即为最优匹 配点。获得匹配点后，根据摄像机透射投影模型恢复图像深度信息。实验结果表明．该方法需处理数据少、计算简便、适于 实时测量。 关键词计算机视觉双目测距匹配矩阵特征提取竞争机制 文献标识码A 文章编号1002—8331一(2006)01—0083—03中图分类号TPl8 Real-timeBinocular and Measurement MatchingDepth Basedon Schema Competing ChaoTian Lu Xiaofang Xiong Qiyong (ElectronicEngneeringDepartment，Fudan 200433) University，Shanghai obstacle has in and from Abstract：vision—based avoidingsystemsuperioritiesdetectingscopetargetinformation．Depth stereois anactiveresearchareainthe vision the is the stereo field，and computer keyproblem matchingprocess．A methodbasedona schemais is ontoa each matching competing proposed．Thematchingprocessmapped matrix，where element a of elements toeachotheras rulesandthe representspair possiblematchingpixels．These compete matching finalwinneris thebest that，the informationcanbe recovered tothe matchingpoints．After depth easily according cameratransmission model．The resultsshowthatless datais neededto projection experiment image process，the is andthe methodisfitforreal-timemeasurement． proceduresimple proposed