The Algorithm Of Image Processing Based On Obstacle Detection In Front Of Train

With the coming of High-speed Era, the rapid increase of train speed and traffic has led to the rise in train accidents. The traditional way that the rail operators detect obstacles with their visual sense is unable to adapt to the present situation. The technology based on computer vision provides an important method for the detection of the iron obstacles.We study a detection method with images for the obstacle in front of the train, which combines with the railway characteristics, provides some auxiliary driving information to train drivers, in order to ensure the safety of train traffic. This article is based on the analysis of images obtained by a single camera, and it studies obstacle detection technology in front of the train.For the detection of static obstacle in a straight line, this article first preprocesses the image to build the image window by the edge of the tracks, so as to narrow the search range of the static obstacles. Then, with the gray histogram, the texture and integrity of the track, we can judge whether there is any obstacle in a straight line. A large number of simulation experiments show that the proposed method can detect static obstacle in front of train in a straight line.For the detection of moving obstacle in a straight line, this article uses the method of optical flow field for testing. Considering the shake and movement of practical camera could affect optical flow field, we compensate for jitter and estimate the motion.The judgment of obstruction on curve is a flawed for vehicle detection system. To compensate this defect, this article uses the following ideas:Firstly, a camera (with wireless transmission equipment) is fixed in the corner,, When the train with super-far focal length camera is moving forward, identifying the bend in the far distance in front of a corner, and immediately starts to receive the wireless image, then uses the algorithm to processed the received image to get the information of the obstacle in front. As the camera is fixed, we use two differential frame methods to judge the obstacles. Finally, by the idea of stationary obstacles positioning in the straight line, the obstacle is positioned on the corners. The experimental results show that this method is effective on the obstacle in the corner.The distance of obstacles from the train is also very important information. At first we introduce a more precise monocular rangefinder model, and then we confirm the model is accurate and reliable with a large number of experiments.