王红星, 徐婉琳, 张勃阳.一种基于改进ORB和PROSAC特征点匹配的V-SLAM算法[J].河南理工大学学报（自然科学版）,2023,42(1):152-159.

WANG H X, XU W L, ZHANG B Y.V-SLAM algorithm based on improved ORB and PROSAC feature point matching[J].Journal of Henan Polytechnic University(Natural Science) ,2023,42(1):152-159.

一种基于改进ORB和PROSAC特征点匹配的V-SLAM算法

王红星1, 徐婉琳1, 张勃阳2

1.河南理工大学 物理与电子信息学院，河南 焦作  454000；2.河南理工大学 土木工程学院，河南 焦作  454000

摘要:随着机器视觉的高速发展，视觉同步定位与地图构建（visual simultaneous localization and mapping，V-SLAM）成为室内定位、导航应用的研究热点。针对传统ORB算法提取特征点分布不均匀的问题，在前端采用四叉树算法管理特征点，实现特征点均匀化分布，并结合渐进抽样一致性（progressive sample consensus，PROSAC）算法剔除误匹配特征点；在后端，采用构建词袋（bag of words，BoW）法对关键帧进行回环检测，判断帧与帧之间是否存在回环，并采用光束平差（bundle adjustment，BA）法进行相机位姿优化，修正相机位姿。在图像特征点提取和匹配实验中，通过与传统ORB算法及其他方法对比，证明本文算法具有较好的运算效率。与ORB_SLAM-modified算法进行轨迹对比实验，分析生成的点云图，结果表明，本文算法具有较高的可靠性和精确度。

关键词:改进ORB特征;渐进抽样一致性;特征点匹配;V-SLAM;3D点云构图

doi:10.16186/j.cnki.1673-9787.2021040137

基金项目:国家自然科学基金资助项目（41807209）

收稿日期:2021/04/30

修回日期:2021/06/09

出版日期:2023/01/25

V-SLAM algorithm based on improved ORB and PROSAC feature point matching

WANG Hongxing1, XU Wanlin1, ZHANG Boyang2

1.School of Physics and Electronic Information，Henan Polytechnic University，Jiaozuo  454000，Henan，China；2.School of Civil Engineering，Henan Polytechnic University，Jiaozuo  454000，Henan，China

Abstract:With the rapid development of machine vision，visual simultaneous localization and mapping （V-SLAM） has become a research hotspot of indoor positioning and navigation applications.Aiming at the non-uniform distribution of feature points extracted by the traditional ORB algorithm，the quadtree algorithm was used to uniformly distribute feature points at the front-end，and progressive sample consensus （PROSAC） algorithm was used to eliminate mismatched feature points.At the back-end，bag of words （BoW） was used to detect loop closure in the key frames and the bundle adjustment （BA） was used to correct the camera position-posture data.Compared with the traditional ORB algorithm and other methods in the experiment of image feature point extraction and matching，it was shown that the proposed algorithm had better efficiency.The trajectory comparison experiment with ORB_SLAM-modified algorithm and the analysis of the point cloud map showed that the proposed algorithm had higher reliability and accuracy.

Key words:improved ORB feature;progressive sample consensus;feature point matching;V-SLAM;3D point cloud composition

 018_2021040137_王红星_H.pdf