Research On The Application Of RGB-D Perceptual 3D Point Cloud Processing In Robot Path Planning

With the continuous development of mobile robot visual navigation technology,the use of color images and depth images provided by low-cost RGB-D sensors to guide the path planning and environment perception of mobile robots has become one of the research hotspots at home and abroad.However,because of the large amount of RGBD information,limited resolution,and weak anti-interference ability,how to effectively and accurately establish a three-dimensional map of the unknown environment under the limited conditions of using only RGB-D without using other sensors,and use it for The stable,efficient and reliable environment perception and path planning of mobile robots is an urgent problem to be solved.Firstly,this dissertation systematically studies the autonomous perception and path planning technology of mobile robots based on RGB-D in unknown environments.When the mobile robot is navigating autonomously under RGB-D perception,the synchronous positioning and map construction algorithm constructs a threedimensional point cloud map of obstacles in the path,analyzes the angle of the point cloud plane,calculates the height of obstacles in real time and judge whether it is passable.Secondly,the original point cloud data output by the RGB-D camera will cause a large amount of noise in the collected point cloud data due to the accuracy error of the equipment and the mutual obstruction of the field of view between environmental objects,which will seriously affect the accuracy of subsequent segmentation and filtering processing is required.Thirdly,this paper uses the ICP method in the TUM RGB-D dataset to perform ORB-SLAM2 based point cloud registration and compares the completeness of map reconstruction with the RGB-D SLAM algorithm,and then analyzes and compares the differences in statistical filtering in the Table Scene dataset.The denoising effect of the parameters,the best parameter K=20,?=2 is obtained.The mobile robot uses the ORBSLAM2 algorithm to construct a point cloud map in an outdoor environment using the ICP method,and then performs passthrough filtering,voxel filtering,statistical filtering and plane segmentation respectively,calculating the slope angle,and implementing motion planning.Finally,the experimental results show that the ICP method used in the TUM RGBD data set of the Technical University of Munich can be applied to ORB-SLAM2.While generating the original sparse feature map,it also generates a point cloud map for mobile robot path planning.,And the improved algorithm is better than the RGB-D SLAM algorithm in map accuracy and positioning accuracy;the optimal statistical filter parameters obtained in the Stanford University's three-dimensional point cloud dataset Table Scene can be applied to outdoor environments,and mobile robots can calculate the size of the front slope,and automatically plan the path according to the calculation result to complete the specified task.