视差引导下的影像线段特征匹配算法  

Image Segment feature matching algorithm with disparity guidance

作　　者：王海彬[1] 吴军[1] 彭智勇[1] 

机构地区：[1]桂林电子科技大学电子工程与自动化学院,广西桂林541004

出　　处：《桂林电子科技大学学报》2016年第2期129-135,共7页Journal of Guilin University of Electronic Technology

基　　金：国家自然科学基金(60962003);广西自然科学基金(2014GXNSFAA118302);广西自动检测技术与仪器重点实验室基金(YQ15108);广西教育厅科研项目(KY2015YB111)

摘　　要：为提高线段特征匹配的效率和准确率,通过引入SGM算法生成立体像对的初始视差信息,改进了影像线段特征匹配算法。该算法利用生成的视差信息建立待匹配线段的梯形搜索窗口,根据线段提供的几何和灰度特性,剔除梯形搜索窗口内的歧义线段,通过最小视差差分准则,迭代计算待匹配线段与搜索窗口内候选线段的匹配代价,从而确定最终的同名线段。实验结果表明,该算法提高了线段特征匹配的效率,且准确率达到98%以上。In order to improve the efficiency and accuracy of segment feature matching, SGM dense matching is used to obtain initial disparity information, segment feature matching algorithm for image is improved. Trapezoidal searching window is established by disparity information. After that, ambiguity segment in this search window are eliminated through geometric and gradation characteristics. Finally according to the minimizing parallax difference, the matching cost between candidate segment and matched segment is calculated iteratively for obtaining the same name segment. The experimental results show that the accuracy rate of segment feature matching is more than 98%.

关 键 词：同名直线 视差信息 梯形搜索窗口 迭代准则 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]