基于改进ORB特征匹配的单目视觉—惯性SLAM算法  

Monocular Visual-inertial SLAM Algorithm Based on the Improved ORB Feature Matching

作　　者：祝晓轩 杨杰[1] 胡继港 ZHU Xiao-xuan;YANG Jie;HU Ji-gang(College of Mechanical and Electrical Engineering,Qingdao University,Qingdao 266071,China)

机构地区：[1]青岛大学机电工程学院,青岛266071

出　　处：《青岛大学学报（自然科学版）》2023年第1期60-64,70,共6页Journal of Qingdao University(Natural Science Edition)

摘　　要：针对VINS-Mono算法在弱纹理环境下运行不稳定、累积误差大等问题,提出了改进ORB特征匹配筛选的单目视觉—惯性SLAM算法。测量数据预处理环节采用了ORB特征提取,在特征点匹配时,使用双向匹配过滤和最小匹配点距离倍数判别剔除误匹配,最后利用随机采样一致性算法完成特征匹配。实验结果表明,与原算法相比,改进VINS-Mono算法的精度较高。Aiming at the problems of unstable operation and large accumulated error of VINS-Mono algorithm in weak texture environment,a monocular visual-inertial SLAM algorithm based on improved ORB feature matching and filtering was proposed.In the preprocessing of the measurement data,ORB feature extraction algorithm was adopted.In the feature point matching,two-way matching filtering and minimum matching point distance multiple discrimination were used to eliminate false matching.Finally,RANSAC was used to complete feature matching.The experimental results show that the improved VINS-Mono algorithm has improved accuracy compared to the original algorithm.

关 键 词：SLAM 特征点提取 特征点匹配 匹配点筛选 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]