【作者】 赵杰； 彭京； 李戈； 王晨；

【Author】 Zhao Jie Peng Jing Li Ge Wang Chen Robot Research Institute, Harbin Institute of Technology, Harbin 150001, China

【机构】 哈尔滨工业大学机器人研究所；

【摘要】 立体视觉匹配技术是计算机视觉领域中的热点和难点。针对以往的区域匹配计算量大而特征匹配算法复杂的缺点,本文提出一种粗细结合的匹配方法,采用区域匹配和特征匹配相结合的方法同时使用两级阈值判断相关函数值实现双目立体视觉的匹配。在匹配的第一阶段实现粗匹配,通过设置两级阈值得到最佳匹配点和候选匹配点;在匹配的第二阶段实现细匹配,利用视差相容原理对粗匹配阶段得到的候选匹配点进行唯一性匹配.实验证明,该方法降低了歧义匹配概率,提高了匹配的灵活性和可靠性。更多还原

【Abstract】 Stereovision matching is focus and nodus in stereovision field. Due to big cost of calculation in area-based matching and complexity of algorithm in feature-based matching, this paper presents a coarse and fine combination matching method. This method uses coarse and fine combination method and two-threshold-value judgment. In the first stage of coarse matching, optimal match points and candidate match points are gained through setting two-threshold-value. In the second stage of fine matching, the unique matching for candidate points is done with disparity correlation. Proved by experiments, using this method increases flexibility and reliability of matching and decreases error-matching probability.更多还原