摘要
针对特征点匹配中误匹配筛选算法鲁棒性差及效率低等问题,提出了一种快速的误匹配筛选算法。文中将基于欧氏距离的大津法与随机取样一致算法相融合,针对传统方法筛选数量问题,在欧氏距离的基础上引入大津法,通过大津法对匹配对进行误匹配剔除,并对数据进行排序以提高算法运行速度,最后利用RANSAC对优化数据进行第三次筛选。实验表明:此算法的准确率要优于主流筛选算法,筛选结果的正确率在95%以上,具有良好的鲁棒性;在运行速度方面,此算法比传统的RANSAC算法减少了80%以上的运行时间,具有良好的实时性。
Aiming at the poor robustness and low efficiency of the mismatch screening algorithm in feature point matching,a fast mismatch calculation and selection algorithm is proposed.This paper combines the Otsu method based on the Euclidean distance and the random sample consensus(RANSAC)methods.Aiming at the traditional method of screening the number problem,the Otsu method is introduced on the basis of the Euclidean distance.The data is sorted to improve the speed of the algorithm,and finally the optimized data is screened for the third time using RANSAC.Experiments show that the accuracy of this algorithm is better than that of mainstream screening algorithms,and the accuracy of screening results is above 95%,which has good robustness;in terms of running speed,this algorithm is 80%less than the traditional RANSAC algorithm.The above running time has good real-time performance.
作者
宋孟良
张国伟
卢秋红
张苏苏
SONG Mengliang;ZHANG Guowei;LU Qiuhong;ZHANG Susu(School of Automation Engineering,Shanghai University of Electric Power,Shanghai 200082,China;Shanghai HRSTEK Co.,Ltd.,Shanghai 201100,China;Shanghai Solar Energy Engineering Technology Research Center Co.,Ltd.,Shanghai 201100,China)
出处
《机械工程师》
2021年第3期30-32,共3页
Mechanical Engineer
基金
上海市2018年度科技型中小企业技术创新资金项目(18CT24H4700)
闵行区2018年度中小企业技术创新计划项目(2018MH037)
2019闵行区产学研项目(2019MHC083)。