Research On Obstacle Detection Method In Front Of The Train Based On Machine Vision

Up to now, China’s railway has undergone six overall great speed rising, the train traffic safety issue has been received more and more attention. In recenr years, with the increasing of passenger and freight transport volume of our country’s railway industry, the occurrence frequency of outside rail road casualties are more and more high. The detection of obstacles in front of the train has very important significance to ensure the safe operation of trains. Using machine vision technology to detect obstacles in front of the train has great practical significance.Aimed at the problems of static obstacle detection and dynamic obstacle detection and tracking, the theoretical analysis and experimental verification are carried out. The system software for obstacle detection in front of the train based on machine vision is designed.In the detection method of static obstacles, firstly, image preprocessing to the collected video images is realized. Then, the tracks contour is extracted by using the chain code tracking algorithm, which is based on the track’s own characteristics, and the detection window is established. Finally, the static obstacles in the detection window are detected through the method of combing image texture features and gray value variance and black-white pixel ratio. The establishment of detection window can effectively reduce the detection range of static obstacles and the computational complexity of detection algorithm in a large part. Experimental results show that the algorithm can achieve real-time detection of static obstacles accurately.The moving targets are detected by using the method of combining three frame differential method and optical flow algorithm through analysing the commonly used moving object detection method, combing the physical environment of train operation and considering the complex changeable of light and so on. Firstly, the moving objects’ initialization is realized by using the three frame differential method. Then, the moving background is removed through the optical flow algorithm. Finally, the actual moving targets are detected.On one hand, the algorithm can solve the large computational complexity problem of optical flow method, on the other hand, which can effectively compensate the illumination changes and achieve accurate detection of moving targets. Using the method of combing image edge direction histograms and Kalman filtering to real-time track the detected moving targets, the proposed algorithm can not only achieve the accurate tracking of single target object, but also realize the effective tracking of multiple objects under the condition of shade.On the basis of machine vision, the system software of detecting obatacles in front of the train is designed, which has functions of video capturing, video storaging, video playback,static obstacle detection, dynamic obstacle detection and tracking and voice alarming and so on. The designed system software can complete the basic monitoring functions.