Research On Detection And Recognition Technology Of Railway Signal Light Based On Machine Vision

With the speed increase of railway transportation and the increase of transportation volume,railway transportation safety has become an extremely important issue.There are many kinds of railway safety concealment,one of which is the misidentification of the signal light.The traditional identification method relies on the human eye to identify the signal light.At present,with the speed increase of the train multiple times,when the speed reaches a certain level,the human eye observation is prone to misjudgment,and the driver may not be able to make corresponding operations in time and accurately.And there will be certain security risks.With the continuous development of machine vision,image processing technology is gradually applied to various fields of industry,and its application to the railway transportation industry can reduce errors caused by human eye observation and reduce the occurrence of railway transportation accidents.Therefore,a model of railway signal detection and recognition based on machine vision is proposed in this paper.Firstly,we compare the four color spaces of RGB,Lab,HSI and HSV,and the five edge operators of Roberts operator,Sobel operator,Prewitt operator,LOG operator and Canny operator.Then,for the image of railway signal lights in complex background,we choose the appropriate color space and edge operator from these color spaces and several edge operators.Secondly,we use the corrosion expansion and Hough transform to extract the geometric characteristics of the railway signal light,so as to obtain a clear edge image of the railway signal.Finally,a recognition model is proposed.The identification step of the recognition model is as follows: firstly,the color space of the railway signal image is converted from RGB to the HSV color space;then it extracts the geometric features of the edge of the railway signal image using techniques such as Canny operator,corrosion expansion,and opening and closing operations;It uses the Hough transform to detect the edge of the round of the railway signal light,and locates the circular shape of the signal light.Finally,it extracts the color of the positioned signal light and recognizes the color of the signal light to interpret the interpretation of the signal light.Experiments show that the recognition model can effectively and accurately identify the color of the railway signal light image.In order to explore a new feasible method,this paper also proposes a deep learning-based method,which uses the tensorflow to train the simulated railway signal image data,and finally generates a CNN-based image recognition model.Experiments show that the CNN-based recognition model is accurate and effective.Based on machine vision,the railway signal light detection and recognition system mainly realizes three functional modules: picture recognition module,video recognition module and analog line control module.It uses the recognition model to realize the identification of the railway signal light and the identification of the video.It calls the Baidu map API to realize simple railway line simulation.The system can effectively identify the color of the railway signal lights and can easily simulate the railway travel route on the map.