Research And Application Of Detection Method Of Vanishing Point In Coal Mine Roadway Scene

In the image,the location of the vanishing point is of great significance to the determination of road direction,approximate azimuth and other information.At the same time,the vanishing point can also be used as a conditional constraint or one of the main calibration standards for detecting the road and the feasible area of the robot.Due to the complexity,closure,and particularity of underground coal mine roadways,the existing robot navigation system is not suitable for this kind of unstructured and complex environment.In this case,the vanishing point in the scene image of the coal mine roadway can be detected to assist the robot to complete visual navigation.However,there are still many difficulties in vanishing point detection in the complex environment of coal mine roadways.Therefore,this paper uses coal mine roadway inspection robots equipped with high-definition cameras to collect roadway scene images,and combines digital image processing technology and image motion estimation methods to complete the following jobs:(1)A method of removing shadow linesAiming at the problem that there are many interference lines in the complex scene images of coal mine roadways,this article first analyzes the causes of these interference lines,and eliminates the shorter interference lines by adding a length threshold.Then,based on the characteristics of the digital image,this article design a method to quickly measure the gradient value of a straight line in an image,and use this method to eliminate the interference straight line caused by the shadow area in the image.Experimental results show that this method can accurately eliminate most of the interfering lines caused by shadow areas in the image.(2)A vanishing point detection method combined with motion estimationIn the coal mine tunnel scene,the lighting conditions are poor,coupled with the rough construction and the accumulation of dust,the texture and linear features in the image are not obvious,and there is a lack of usable structured information for vanishing point detection.To solve this problem,this article combines motion estimation technology to make full use of the relationship between consecutive frames of images,and introduces a scaling factor on the basis of the traditional block matching algorithm,so as to find the straight line of the image block motion trajectory and use it to assist in vanishing points detection.In the process of vanishing point detection,after converting the detected straight line into sample points in the parameter space,in order to reduce the influence of abnormal sample points and improve the accuracy of vanishing point detection,this paper uses the LOF algorithm to calculate the abnormal factor value of each sample point.The abnormal factor value of the sample point and the length of the corresponding line are used as the standard to measure the importance of the sample point,and the weight function of the weighted regression algorithm is designed for the regression estimation of the vanishing point.Experimental results show that this method can more accurately detect vanishing points in the scene images of coal mine tunnels.(3)Method and application of vanishing point detection based on improved RANSAC algorithmIn order to further reduce the number of interference lines in the coal mine tunnel scene and improve the accuracy of vanishing point detection,this paper firstly uses FCN network to complete the semantic segmentation of the image,thereby obtaining the region of interest(referring to the wall),reducing the source of interference,and then extract linear features in the region of interest.In the process of vanishing point estimation,this paper improves the traditional RANSAC algorithm,and assigns a quantifiable standard to measure its importance to each interior point passing through the line.By calculating the weight of all interior points on the line to judge whether the straight line is the best straight line,and get accurate vanishing point coordinates based on the best straight line.Finally,this paper takes the detected vanishing point as a constraint condition to complete the recognition of the road surface area in the coal mine roadway scene image.The experimental results show that the detection accuracy of the vanishing point of this method has been further improved,and the vanishing point is also used to identify the road area.This paper realizes the vanishing point detection in the complex scene of the coal mine roadway,and successfully applies it to the recognition of the road surface area,which has certain guiding significance for the coal mine roadway robot to realize the visual navigation.The thesis includes 66 figures,6 tables and 74 references.