Research On Path Planning Of Mine Rescue Robot Based On Machine Vision

The coal mine environment will become extremely complex after accidents such as gas explosions,roof falling and fires.The rescue task will become quite difficult because of the occurrence of the secondary accidents,the harmful gases and excessive local temperatures.In order to reduce casualties and economic losses caused by coal mine accidents during the rescue process,this paper will be focused on the study of path planning of rescue robot which is based on the machine vision in tunnel environment.The specific work is as follows:(1)The calibration of binocular system for mine rescue robot is studied.First of all,using the zhangzhengyou calibration method,the internal parameters and the relative position between the left and right cameras are got by using MATLAB.Then,aiming at the problem that the luminance value of left and right images are not equal,a calibration method based on binocular vision system is proposed.The method calibrates the consistency of the luminance values of the left and right cameras by calibrating the camera's response function and optical vignetting.It lays the foundation for edge detection and stereo matching.(2)The recognition of obstacles in tunnel scene image is studied.Firstly,the wavelet threshold de-noising algorithm is used to reduce the noise of the tunnel scene image,and the edge detection of the image is carried out by using the Canny operator after the de-noising.So the extraction of the edge information of the mine environment is realized.Then the image is segmented quickly by the watershed segmentation algorithm.For the over segmentation of image,we use K-means clustering algorithm to cluster.After clustering,the Graph Cut algorithm is used to extract foreground and get the two-dimensional regional features of obstacles.Finally,we use the SURF stereo matching algorithm to get the disparity map of the left and right images,and achieve the distance between obstacles and rescue robot.(3)The path planning of mine rescue robot is studied.For the path planning of static obstacle areas,aiming at the problem of goal no reachable with obstacle nearby(GNRON)and local minimum in traditional artificial potential field method,the improved algorithm is studied in this paper.We have added a modified repulsive factor to the repulsion function and set up a temporary target point,so that the rescue robot can complete the task of avoiding static obstacles.For the path planning of dynamic obstacle areas,the artificial potential field method and the rolling window method are fused in this paper.The rescue robot can set the waiting time under indirectly passing state,so that it can evade the dynamic obstacles and complete the path planning.