Research On Visual SLAM Algorithm Of Mobile Robot Based On Depth Camera

With the rapid development of the national economy and science and technology,mobile robots are playing an increasingly important role in the fields of home services,industrial manufacturing and intelligent logistics.When the mobile robot is performing tasks in an unknown environment,the robot needs to be able to accurately locate and efficiently plan paths.SLAM is a feasible solution,but the traditional SLAM algorithm still has deficiencies in feature point extraction and path planning.In order to improve efficiency and reduce the amount of calculation,this paper proposes an improved SLAM algorithm for mobile robot vision based on depth camera.The improved algorithm improves the efficiency of feature point extraction and matching,and reduces the amount of calculation required for path planning..First of all,this research is based on the vision field of depth cameras.An algorithm for keypoint detection and matching is proposed,which is called HHFT(Hue Histogram Feature Transform).The contents of HHFT are: firstly outline several traditional keypoint detection algorithms and point out their advantages and disadvantages.Image segmentation is based on the color gradient image by applying watershed transform to divide the image into "ridge" and "catchment basin".Because keypoint is located in the "ridge",the area of keypoint extraction is reduced,the number of keypoints is reduced.After obtaining the coordinate of the keypoints,the hue histogram of pixels within a certain range around them is used for matching.After the matching process,the results are further optimized using the RANSAC algorithm.The final experimental results prove that the HHFT takes less time and has better performance of robustness in scale and perspective changes.From the perspective of safety and efficiency of path planning,improvement measure is proposed for the RRT(Rapidly Exploring Random Tree)algorithm,which was named RRT-Rectangular method.It is based on the RRT,maps are divided into feasible grids and obstacle grids.Use the feasible grid where the starting point is as the initial grid,and select the next feasible grid in turn according to certain rules.Turning the path planning problem into a raster queue selection problem.Finally,connect the center points of the feasible grid to obtain a feasible path.The Experimental results show that compared with the traditional RRT algorithm,the path length of the improved algorithm is reduced by 5.82%,and there is no risk of collision.Finally,based on the ROS system and MATLAB simulation platform,the HHFT and RRT-Rectangular algorithms are implemented respectively.And the calculation results of time consumption and other aspects are given.By intuitively comparing with the results of classical algorithms,the effectiveness and feasibility of the algorithm designed in this paper are verified.