Research On Depth Recovery Based On Stereo Vision And Path Planning Of Mobile Robot

Nowadays,artificial intelligence has become a hot area in the field of scientific research.Intelligent robots begin to enter people's lives slowly.Robots also become more sophisticated and intelligent from their clumsy shapes and simple functions.Among them,the mobile robot is a very effective transportation and transport tool,which is widely used in various fields.How to make robots have human like eyes and automatic obstacle avoidance planning route is a hot spot in the field of mobile robot research.In this paper,the core algorithms of the two problems of binocular stereo vision and automatic obstacle avoidance planning route are studied.For depth recovery for binocular stereo vision,a stereo matching algorithm based on hybrid robust matching cost and improved guided filter is proposed.On the cost calculation step,a new matching cost with robustness to illumination distortion is proposed,which combines the color and gradient information of the image.On the aggregation step,an improved guided filter is proposed and it is used to aggregate the cost.Compared with the traditional guided filter the improved guided filter can preserve the edge details better and improve the edge details of disparity map.On the post processing step,a new post processing method is proposed.First,consistency checking method is used to mark inconsistent disparity values and wrong disparity values.Then a fast implementation of K-means clustering method is used to segment the lightness values of left image pixels.And then based on the segmentation image and marked disparity map,the disparity map is reconstructed.Experimental results on the Middlebury stereo vision dataset show that the proposed method has an average error of 5.29%.The proposed algorithm is robust to illumination distortion and noise,and it has low computational complexity.For path planning,an improved RRT-based motion planning is proposed for the nonholonomic mobile robots.The environmental constraints and the vehicle's own constraints are considered in this algorithm.The fermat point of triangular geometry is introduced into the RRT algorithm.The pruning function is applied in post-processing path to remove the needless nodes that the path length can be shortened.The fixing angle intersection method is proposed to improve the turning path smoothness.Finally,the Cantmull-Rom interpolation is used to improve the path smoothness.In this paper,the improved RRT algorithm is compared with the basic RRT algorithm and BI-RRT algorithm.The results of simulation show that the convergence rate of the improved RRT algorithm is faster than the other two,and the smooth path with the shortest distance can be produced which can also satisfy the nonholonomic constraint of mobile robots.