一种点线特征融合的双目同时定位与地图构建方法  

A Stereo Simultaneous Localization and Mapping Method Based on Point-line Feature Fusion

作　　者：蒋林[1] 夏旭洪 韩璐 邱存勇[1] 张泰 宋杰 JIANG Lin;XIA Xu-hong;HAN Lu;QIU Cun-yong;ZHANG Tai;SONG Jie(School of Electrical Engineering and Information, Southwest Petroleum University, Chengdu 610500, China;Huabei Oil Communication Co., Ltd. of China National Petroleum Corporation, Renqiu 062550, China)

机构地区：[1]西南石油大学电气信息学院,成都610500 [2]中国石油华北油田通信有限公司,任丘062550

出　　处：《科学技术与工程》2020年第12期4787-4792,共6页Science Technology and Engineering

基　　金：国家自然科学基金青年基金(51607151);四川省创新创业资助项目(S201910615055)。

摘　　要：基于特征点的视觉同时定位与构图方法依赖于图像质量以及可提取的特征点数量,且稀疏的特征点不能直观表达环境的结构信息。为此,提出一种将图像的点特征和线段特征融合的双目同时定位与构图方法。算法前端提取图像的点特征和线段特征,进行特征跟踪并完成相机位姿求解,从跟踪线程中分离出特征提取线程,进一步提升了前端线程的帧率。后端采用集束调整对局部地图进行优化,利用基于词袋模型的闭环检测以抑制系统的累积误差。最后结合点线特征共同构建环境地图。在公开数据集上进行了实验,与当前主流算法相比,提出的算法在保证定位精度的同时能够获得更丰富的环境地图,具备较好的鲁棒性与实时性。Feature-based visual Simultaneous Localization and Mapping methods depend on the image quality and the number of feature points that can be extracted,while the sparse feature points cannot visually express the structural information of the environment.Therefore,a stereo visual simultaneous localization and mapping(SLAM)method combining point features and line segment features was proposed.In the front end of the algorithm,the point features and line segment features were extracted,and then feature tracking was performing to solve the camera pose.The feature extraction thread was separated from the trace thread,which further improved the frame rate of the front-end thread.In the back end,bundle adjustment was used to optimize the local mapping.And a closed-loop detection method based on bag-of-words model was to suppress the accumulated error of the system.Finally,the environment map was constructed by combining with the points-lines.Compared with the current mainstream algorithms,experiments on the public dataset show that the proposed SLAM algorithm can obtain a richer environment map with a great positioning accuracy and good robustness.

关 键 词：同时定位与构图 点线特征融合 图优化 集束调整 闭环检测 

分 类 号：TP391.11[自动化与计算机技术—计算机应用技术]