Research On Simultaneous Localization And Mapping Using Point And Line Features

Simultaneous Localization and Mapping(SLAM)is a key technology for robots to implement automatic navigation.The purpose is to locate their own location in an unknown environment and build the surrounding environment for navigation.For Visual Simultaneous Localization and Mapping(vSLAM),there are two challenges: one is robust operation in texture missing or artificial environment,and the other is the inaccuracy of location caused by motion blurring and illumination changes.In order to solve these challenges and introduce line features to the environment,a vSLAM based on point-line fusion is proposed.A complete scheme based on point-line fusion in the field of RGBD SLAM is proposed,including tracking,local mapping and loop closure detection.In the tracking algorithm,the line segment extracted by improving the existing line segment extraction algorithm,and the line segment is matched according to the word bag model,which solves the problem that the real-time tracking rate of the vSLAM based on the line feature is low.In the local mapping and loop closure detection algorithm,the Plucker coordinates are used to express the spatial line and define the re-projection error of the line,so that the back-end optimization error model of the point-line fusion can be unified.The use of orthogonal representations to optimize spatial straight lines is used to solve the problem of straight-line over-parameterization in optimization and instability in optimization caused by straight-line Plucker constraints.The experimental results using the benchmark data set show that the SLAM proposed in this paper not only improves the accuracy and robustness of pose estimation when only point features are used,but also improves the accuracy of map reconstruction.Although the introduction of line features makes the back-end optimization and feature matching time increase,the proposed method can reach 25 fps due to the efficient line segment extraction algorithm and efficient word bag matching algorithm.