Stereo Visual SLAM System Reasonably Use Point and Line Features

Abstract

AbstractTraditional feature-based SLAM systems rely on point features in the environment to recover the camera pose and build an environmental map. With the in-depth research of scholars, in order to make up for the disadvantages of point features in a low texture environment, stereo visual SLAM systems that combine both points and line segments are proposed. Although the stereo visual SLAM systems that combine both point features and line features improve the accuracy of tracking and they also increase the computational burden of the computer and reduce the tracking efficiency. However, in the actual environment, the environment will not always be in a state of low texture, so our work considers that the line feature can be selectively used. We selectively use line features during the tracking process and use line features as a supplement to point features in a low-texture environment. The main content of our work is proposing an analysis method that analyzes the change of the environmental characteristics that have been tracked during the tracking process to make the SLAM system possible to make good use of point features and line features in the tracking process. We test our system on public datasets and compare our results with state-of-the-art methods. The test results show that our stereo visual SLAM method can obtain more accurate results than the stereo visual SLAM system that uses points and even the stereo visual SLAM system that combines both point features and line features.KeywordsStereo visual SLAMPoint featuresLine featuresTracking feature analytical method