改进边缘能量聚集的半全局匹配算法  

Semi-global Matching Algorithm Based on Improved Energy Aggregation on Edge

作　　者：孙辉 李浩[1,2] 陈智君[1] 

机构地区：[1]河海大学地球科学与工程学院,江苏南京211100 [2]岩土钻掘与防护教育部工程研究中心,湖北武汉430074

出　　处：《地理空间信息》2017年第11期80-83,共4页Geospatial Information

基　　金：国家自然科学基金资助项目(41471276);工程地质与岩土防护学术创新基地岩土钻掘与防护教育部工程研究中心开放研究基金资助项目(201508);水利部黄土高原水土流失过程与控制重点实验室开放课题基金资助项目(2015001)

摘　　要：针对传统半全局算法匹配时间长和视差突变处易存在误匹配的问题,提出了一种改进边缘能量聚集的半全局匹配算法。在半全局匹配算法的基础上,引入金字塔分层策略,通过对图像降采样进行加速匹配,降低匹配算法的复杂度,减少内存的消耗。利用Canny算子获得分层图像的边缘信息,匹配时根据边缘信息改进边缘处的代价聚合并构建自适应窗口减小视差搜索域,获得正确率更高的视差图。结果表明,该算法可将明显提高匹配速率,降低运算复杂度,提高正确率。As to the problem of the long time and the inaccuracy at abrupt changes of disparity, this paper presented a semi-global matching algorithm based on improved energy aggregation on edge. Based on the semi-global matching algorithm, the paper introduced the pyramid layering strategy and accelerated the matching by the downscale images to reduce the complexity and memory consumption. This algorithm used Canny operator to obtain the edge information of layered images. According to the edge information, the algorithm improved the cost aggregation at the edge and structured adaptive window to reduce the disparity search range for a high-accuracy disparity map. The results show that the algorithm can significantly improve the matching rate, reduce the computational complexity and improve the accuracy.

关 键 词：半全局匹配 CANNY算子 边缘检测 自适应窗口 

分 类 号：P208[天文地球—地图制图学与地理信息工程]