Application Research Of Image Processing Technologies On Safety Protection For Shunting Operation

When the locomotive for shunting operation is output and input of storage, it needs locomotive drivers to outlook and confirm shunting signal, even some locations rely entirely on watching and confirming signal by locomotive drivers. It can increase the labor intensity of drivers if locomotive drivers carrying on s hunting operation for a long time, and it is possible to cause of shunting accidents by human factors. Therefore, it is of importance practical significance that the image processing technologies are applied to shunting operation safety protection to build a video surveillance system. Using this system, the signal detection and recognition of the shunting places can be carried out automatically and offers relevant information, the locomotive drivers can make judgment through these information.According to the shape of the shunting signal lamp and the light has the characteristics of the circular, and in view of the complexity of shunting work environment, a method of improved randomized Hough transform shunting signal detection is adopted. The front images of locomotives running that are taken are about the track way of the locomotives if the camera equipment angles of locomotives are installed appropriately in practical application, so the signal lamp that is the nearest distance from this track way is considered as t he detection result. The experimental results show that this method can efficiently reduce complexity of image process, and improve the precision of algorithm detecting.The SVM is used to classify the color of the signal lamp after the signal lamp detection. The color features of models in color space are selected as the identification features, the color features of signal lamps detection areas are considered as the input of SVM classifier, and the color of signal lamps are considered as the output, the color recognition of signal lamps is implemented through above operation. The SVM classifier is off-line trained by all kinds of sample images before the real-time acquisition signal lamps images collected are recognized by SVM classifier, it can reduce time of algorithm processing this way.After the signal lamp recognition, the mathematical relationship between the pixels of the signal images and the practical distance of the signal lamp and the locomotive is got through related ranging technologies, then appropriate ranging method is selected for live working situation. The technologies of shunting signal detection, recognition, ranging are combined with associated hardware equipment to get the color and distance information of signal lamp, which is applied to shunting operation of station at last.