Research On Indoor Mapping And Path Planning Based On Visual SLAM

In recent years,map building and path planning for mobile robots have gained widespread application in fields such as industrial production and smart homes,becoming an integral part of modern life.Simultaneous Localization and Mapping(SLAM)is a critical technology that enables mobile robots to perceive their environment by using cameras or laser range finders,accurately locate themselves,and construct an environment map for navigation and path planning.This study specifically aims to improve the visual navigation of mobile robots in indoor settings by utilizing RGB-D cameras as the visual sensor for perception.It addresses the challenges of low positioning accuracy and incomplete map construction in dynamic environments,as well as the issue of redundant path planning in visual SLAM algorithms.A target detection-based indoor dynamic scene SLAM method is proposed,which filters out dynamic information from the indoor dataset collected by RGB-D camera sensors,thereby improving the positioning accuracy of SLAM.In the tracking thread of ORB-SLAM2,a YOLOv4 target detection thread is added to detect prior dynamic targets in the experimental scene,and then the LK optical flow method is used to filter out dynamic feature points.Finally,the camera pose is estimated using Pn P based on the obtained static feature points for mobile robot localization.Experimental results show that the proposed algorithm significantly improves the inaccurate positioning problem of traditional ORB-SLAM2 systems in indoor dynamic scenes.A method for indoor 3D dense map construction is proposed.Firstly,the collected depth images are preprocessed by bilateral filtering to denoise the depth maps.Secondly,the sparse point clouds constructed by the traditional ORB-SLAM2 system cannot provide map information for mobile robots.Therefore,a dense point cloud mapping thread is added to the local mapping thread to construct a dense point cloud map.Finally,because dense point clouds occupy large memory space and contain much redundant information,a more flexible and compressed octree map is constructed to preserve more valid information,greatly reducing the storage space of the map and better preparing for subsequent path planning.An improved ant colony algorithm for mobile robot path planning is proposed.The ant colony algorithm is a probabilistic algorithm for finding the optimal path in a given problem.Since all parameters in the ant colony algorithm remain constant,the algorithm’s results rely too heavily on the set information pheromone parameters.To improve the allocation of ant colony algorithm parameters and pheromones,the importance factor parameters for both the heuristic factor and the pheromone importance factor are changed in each iteration.This approach utilizes all meaningful parameter spaces as much as possible and effectively improves the efficiency of mobile robot path planning.