Camera Localization And Scene Reconstruction Method Based On Stereo Image Sequence

Camera localization and scene reconstruction are hot topics in the field of computer vision and robotics.Vision-based camera localization and scene reconstruction often suffer from inoperable or degraded accuracy when encountering problems such as motion blur,low texture,or loop closure false detection.In response to this type of phenomenon,this paper has carried out the following three aspects of research.(1)This paper proposes a stereo visual simultaneous localization and mapping algorithm combined with inertial measurement unit.The algorithm extracts and matches the ORB features of the stereo image sequence,then estimates the camera pose by minimizing the projection error of the point feature,and uses the IMU parameters to estimate the camera pose when encountering the motion blur image,so that the algorithm can obtain continuous and reliable camera pose.The algorithm is simulated on the EuRoC MAV dataset.The experimental results show that the algorithm can obtain continuous and complete camera trajectory and achieve reliable localization.(2)This paper proposes a stereo visual SLAM algorithm based on point features and line segment features.The algorithm extracts and matches the ORB point feature and the LSD-LBD line segment feature for the stereo image sequence,estimates the camera pose by minimizing the projection error of the point feature and line segment feature,and uses local bundle adjustment of the point feature and line segment feature to optimize the camera pose and feature 3D coordinates of the keyframe,the algorithm can still achieve reliable camera localization in low-texture scenes.The algorithm is simulated on KITTI,EuRoC MAV and it3 f datasets.The experimental results show that the algorithm can obtain higher precision camera trajectory in low-texture scene and improve the effect of scene reconstruction.(3)In this paper,a stereo visual SLAM algorithm based on loop closure optimization of point features and line segment features is proposed.The algorithm calculates the similarity scores of point features and line segment features between two image frames and weights them,and uses the point features and line segment features to calculate the relative pose between keyframes,so that the visual SLAM algorithm can also comprehensively utilize the point features and line segment features in the loop closure optimization.The algorithm is simulated on KITTI and it3 f datasets.The experimental results show that the algorithm can effectively solve the problem of false loop closure detection,thus improving the accuracy of camera localization and improving the effect of scene reconstruction.