Doors & Windows Status Detection With Open And Close For Boxcar In Railway Based On Computer Vision

With the development of China's economy,the demand for railway freight is increasing.The density and speed of railway freight cars have been greatly improved,which is followed by railway freight safety.Among railway freight cars,boxcars are one of the main models used to transport valuable cargo that is afraid of rain and snow and sun exposure.During the running process,the boxcar doors and windows may be open due to bumpy body and locking structure damage,which poses a serious threat to the safety of cargo transportation.Therefore,it is of great significance to detect the opening and closing state of boxcar doors and windows and alarm in real time.Relying on the railway station video surveillance system,this paper studies the boxcar doors?windows open status detection algorithm,obtaining the images of the both sides of boxcar during the operation of freight cars in real time,to achieve the real-time automatic detection of boxcar doors and windows open status by the use of computer vision technology for image processing.In addition,the detection of the tanker cover and the detection of the damage of the gondola body in the railway wagon are discussed,and the preliminary solutions are put forward.The main research contents and innovations of this paper are as follows:(1)Aiming at various quality problems of boxcar images,image preprocessing method is studied.The image is de-noised by bilateral filtering,and the edge of the image is retained while filtering the noise;the image contrast is enhanced by gray linear transformation.In the process of segmenting the body of boxcar,the threshold segmentation based on gray histogram,the region segmentation based on region growing method and the edge segmentation algorithm based on boxcar structure are studied.After the test and analysis,the edge segmentation algorithm based on the boxcar structure is used to obtain a good body segmentation results.(2)The boxcar doors open status detection algorithm is proposed.According to the characteristics of the line distribution on the door,the door is located by straight line detection and line clustering,and then the opening and closing state of the door is further classified.In the straight line detection,the line detection algorithms based on combination morphology,Hough transform,LSD are studied.After comparison and analysis,a line detection algorithm combined combination morphology and Hough transform is used.The hierarchical clustering algorithm based on the realization of union-find sets is studied to cluster the lines,and the location of the door is realized.Through the experiment,The accuracy of the algorithm to detect door opening faults has reached 95.35%,and meets the requirements in real-time.The single image average processing time is 134 ms.(3)The boxcar windows open status detection algorithm is proposed.According to the spatial relationship between the inner window and the outer window while the window is opened,the inner window and the outer window are detected in order to judge the opening and closing state of the window.Based on the line distribution of the inner window,the algorithm of edge extraction,line detection and cluster analysis is studied to locate the inner window.According to the characteristics of the elliptical structure on the outer window,the ellipse detection algorithms based on the least square method,the Hough transform and the SVM are studied.After the experiment and analysis,the ellipse detection algorithm based on SVM is selected to detect the outer window.In addition,sample preparation,feature extraction and optimization of SVM parameters were studied.Finally,The experiment shows that the accuracy of detecting window failure is 96.97%,and the average processing time of single image is 130 ms.(4)The detection of the tanker cover and the damage of the gondola body in railway wagon are discussed.Through the research and analysis of these two kinds of faults,the corresponding detection scheme has been put forward initially.The image samples used in this study all come from the images of box cars and tank cars actually collected at a certain station in Wuhan.