基于动态自适应调整阈值的SURF改进匹配算法  

Improved SURF Matching Algorithm Based on Dynamic Self-adaptive Adjustment Threshold

作　　者：李珊[1] 李浩[1,2] 余兰[1] 王莎[1] 胡屹群 

机构地区：[1]河海大学地球科学与工程学院,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：《地理空间信息》2017年第3期100-102,共3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)

摘　　要：相比于SIFT算法,在图像匹配速度上SURF算法具有明显优势,但其错误匹配率较高,因而选择合适的距离比阈值可有效提高匹配的准确度。由于影像的纹理结构各不相同,以RANSAC算法优化后的点对结果与SURF直接匹配得到的结果的比值作为动态调整的依据,提出了基于动态自适应调整阈值的SURF改进匹配算法。实验表明,自适应调整的阈值能使匹配正确率提高到95%以上。Compared with SIFT algorithm,SURF algorithm has an obvious advantage in image matching speed.Because of SURF algorithm’s high error matching rate,choosing an appropriate threshold of distance ratio can effectively improve the accuracy of matching.Considering the different texture structures of different images,an improved SURF matching algorithm based on dynamic selfadaptive adjustment threshold was proposed.The ratio of the optimized results of RANSAC algorithm and the direct matching results of SURF algorithm was as the basis of dynamic adjustment.The result shows that the self-adaptive adjustment threshold can increase the matching accuracy to more than 95%.

关 键 词：动态自适应 SURF算法 匹配率 RANSAC算法 距离比 

分 类 号：P231[天文地球—摄影测量与遥感]