Research On Safety Warning System For Station Shunting Signal

The ground shunting signal displays real-time on the shunting locomotive, which has animportance of shunting operation safety protection. Our country station has protective blindzones, when shunting operation and locomotives loading and unloading of storage, and has nosuitable safety protection system for the main of line locomotives temporary shuntingoperation along railway line, mainly relies on the driver outlook signal light condition forshunting operation, which requires the driver to keep highly attention, easy to cause the lossof the driver’s mental fatigue and attention, at the same time, the outdoor work easilyinfluenced by visual angle and the weather, easy to happen shunting accidents caused byhuman factors. Therefore, in order to solve this problem, using machine vision technology incombination with image processing and other related knowledge, station shunting signalreal-time detection, recognition and distance measurement technology are mainly studied.According to the shape of the shunting signal light and the light has the features of thecircular, in view of the complexity work environment of station shunting operation, a methodof railway shunting signal detection is adopted, which is a based on Principal ComponentAnalysis (PCA) and randomized Hough transform shunting signal detection method, andcombining the features of station line to realize accurate detection of this track of shuntingsignal light. Experiments prove that this method can effectively simplify the complex back-ground image, improve the detection speed and has good precision, and the detection ofcolored light has a good effect. On this basis, through the analysis of color space model, thefeatures value H, S, I reflecting the color nature of the display signal light area are extracted asthe input of the modified BP neural network, the display color of shunting signal light as theoutput, BP neural network color classifier is obtained by the offline training, and implementsthe identification of the display status of shunting signal light on line. The experimentalresults show that suitable BP neural network’s parameters are selected, classifier recognitionaccuracy is higher. Meanwhile on the basis of accurate detection of the shunting displaysignal, according to the target distance and target object pixel height in the image has acorresponding relationship, which can be obtained by statistical data, and fitting the functionalrelation for the distance of shunting display signal lights to shunting locomotives and signallight pixels height in image, the experimental results show that the function relation can meetthe needs of the shunting signal light distance measurement.On the basis of the above results, the station shunting signal safety warning system basedon machine vision is designed and implemented, through the display device installed inshunting locomotive, provides the driver with signal light display status and positioninformation in front of the locomotive running, and if the shunting signal is judged as inhibit signal, then the warning information is provided the driver to remind safety. The stationshunting signal safety warning function is completed.