基于ADCensus的改进半全局匹配方法  

Improved Semi Global Matching Method Based on ADCensus

作　　者：李珊[1] 李浩[1,2] 王莎[1] 陈智君[1] 

机构地区：[1]河海大学测绘科学与工程系,江苏南京211100 [2]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,湖北武汉430074

出　　处：《甘肃科学学报》2017年第3期19-23,36,共6页Journal of Gansu Sciences

基　　金：国家自然科学基金(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项目(201508)

摘　　要：考虑到AD相似性测度对灰度信息及纹理丰富的区域的优势和Census变换对图像辐射差异具有鲁棒性的优势,将AD和Census算法融合替代单一匹配代价来改进半全局密集匹配算法,可有效提高算法的匹配精度和抗干扰性。同时用金字塔分层匹配策略来约束视差的搜索范围,实验结果表明,该方法在保证匹配精度的同时,减少内存消耗和运行时间,适用于无人机影像的密集匹配。Taking into account the advantages of the AD similarity measure on the gray information and texture rich region and the advantage of the Census transform on the image radiation difference, the AD and Census algorithm are combined to replace the single matching cost to improve the semi global dense matching algorithm,so matching precision and anti-interference of algorithm is effectively improved.At the same time,the Pyramid layered matching strategy is used to constrain the search range,the result shows that this method can reduce the memory consumption and running time,and is suitable to dense matching of UAV images while it can ensure matching precision.

关 键 词：匹配代价 ADCensus算法 半全局匹配 金字塔分层策略 

分 类 号：P231.2[天文地球—摄影测量与遥感]