摘要
针对随机采样一致算法计算复杂且不能完全消除误匹配的缺点,该文提出了一种特征匹配改进算法来克服该缺点。在特征匹配后根据距离约束筛选匹配点对,然后基于随机采样一致算法进行二次筛选,利用最终的匹配结果构建优化问题,基于图优化理论进行运动估计。匹配筛选实验验证了算法的有效性,结果表明,经本文算法筛选得到的特征匹配,满足单目视觉里程计的数据要求,且轨迹跟踪实验里程计绝对轨迹误差的均方根误差值较现有方法下降了22.66%。
Aiming at the shortcomings of the random sample consensus algorithm which was complex in computation and could not completely eliminate mismatching,an improved feature matching algorithm was proposed to overcome the shortcomings of the algorithm.After feature matching,matching point pairs were screened according to distance constraints,and then second screening was carried out based on random sample consensus algorithm.The final matching results were used to construct optimization problems,and motion estimation was carried out based on graph optimization theory.The validity of the algorithm was validated by matching and screening experiments.The results showed that the feature matching obtained by this algorithm met the data requirements of monocular visual odometry.The results of trajectory tracking experiments showed that the RMSE value of the absolute trajectory error of odometry was 22.66%lower than that of the existing methods.
作者
王俊鹏
陈国良
王睿
靳赛州
WANG Junpeng;CHEN Guoliang;WANG Rui;JIN Saizhou(Key Laboratory for Land Environment and Disaster Monitoring,Ministry of Natural Resources,China University of Mining and Technology,Xuzhou,Jiangsu 221116,China;School of Environment Science and Spatial Informatics,China University of Mining and Technology,Xuzhou,Jiangsu 221116,China)
出处
《测绘科学》
CSCD
北大核心
2020年第6期51-56,共6页
Science of Surveying and Mapping
基金
国家重点研发计划项目(2016YFB0502105)
国家自然科学基金项目(41371423)
江苏省自然科学基金项目(BK20161181)
江苏高校品牌专业建设工程资助项目(PPZY2015B144)。
关键词
单目视觉里程计
特征匹配
随机采样一致算法
距离约束
图优化
monocular visual odometry
feature matching
random sample consensus algorithm
distance constraints
general graphic optimization
