The SLAM Algorithm With The Optimization Module Of Loop Closure Based On The Indoor Dynamic Environment

With the development of unmanned devices,the application of Instant Location and Scene Reconstruction Technology(SLAM)in indoor scenes has attracted more and more researchers' attention.The premise of the application of the traditional SLAM algorithm is that the scene information remains stationary,and the geometrical constraint relationship in the space is used to calculate the pose change of the current sensor.However,when the SLAM algorithm is actually applied to the process of unmanned devices navigating indoors,the scene state information is often difficult to maintain static.This paper mainly studies the interference problem caused by dynamic objects in indoor scenes to SLAM algorithm,and studies the closed-loop detection and mapping process of SLAM algorithm.The main contributions of this paper are as follows:1.Aiming at the problem of introducing interference characteristics into moving objects,it is proposed to eliminate the interference characteristics introduced by dynamic objects by using the magnitude of the change of depth information,and improve the matching precision of closed-loop detection.At the same time,the algorithm only deals with the serious situation of dynamic object interference according to the depth information in the scene.Reduce computational overhead and improve the real-time performance of the algorithm on mobile unmanned devices.2.Construct the PointCloud view for SLAM System:for the occlusion problem in the scene caused by dynamic objects,by combining depth learning and geometric constraints in space,the dynamic objects in the image are segmented by pixels,and the static information in the scene is preserved.The static scene information blocked by the dynamic object is recovered by combining the camera pose information and the static image information corresponding to the plurality of key frames.In this paper,the algorithm is applied to the common dynamic dataset and actual scene of TUM,and the relevant experimental results are given.The experimental results show that the algorithm-optimized closed-loop detection module can reduce the false alarm rate to 11.3%in dynamic scenes,and effectively preserve the static information in the scene by segmenting the dynamic objects in the scene to reconstruct a high-precision 3D point cloud map.It provides precise guarantee for the self-navigation of the unmanned equipment entering the scene.