摘要
针对视觉导航中现有的特征匹配算法召回率低、耗时长的问题,提出一种基于运动平滑性与RANSAC算法结合的图像特征匹配算法。首先将图像网格化,通过运动平滑性约束处理,找出误匹配率低的图像网格区域;然后,利用RANSAC算法计算出图像间近似的单应矩阵;最后,利用单应矩阵对初次匹配结果进行筛选,得到优化后的匹配结果。实验结果表明:所提出的算法在不同类型的图像匹配中,召回率平均提升了7.5%,F值平均提升了5.4%;相比较于传统RANSAC算法,运算时间平均减少了27.8%,适用于一些对实时性要求较高的场景。
Aiming at the problem that the feature matching algorithm in visual navigation has low recall rate and poor real-time performance, an optimized image feature matching algorithm based on motion smoothness and RANSAC algorithm is proposed. Firstly, the images are divided into several grids, and the regions with low mismatch rate are located by using motion smoothness. Then, the homography matrix between images is calculated by using RANSAC algorithm. Finally, the homography matrix is used to filter the initial matching results, and the optimized matching results are obtained. Experiment results show that the recall rate and the F value are improved by an average of 7.5% and 5.4% respectively in different types of image matching. Compared with traditional RANSAC algorithm, the operation time is reduced by 27.8%, which is suitable for the scenes with high real-time requirements.
作者
程向红
李俊杰
CHENG Xianghong;LI Junjie(Key Laboratory of Micro-inertial Instrument and Advanced Navigation Technology,Ministry of Education,Southeast University,Nanjing 210096,China;School of Instrument Science&Engineering,Southeast University,Nanjing 210096,China)
出处
《中国惯性技术学报》
EI
CSCD
北大核心
2019年第6期765-770,共6页
Journal of Chinese Inertial Technology
基金
国家自然基金项目(61773116)
关键词
特征匹配
运动平滑
召回率
单应矩阵
feature matching
motion smoothness
recall rate
homography matrix
