Research Of Rails Turnout Recognition Technology Based On Infrared Thermal Imaging Image Processing

Railway transportation plays a more and more important role in daily life in China.Recently, the requirement of safety is getting much more concern. To ensure awell-functioned railway system, and avoid the situations of crash, railway switches needto be recognized accurately. In this work, infrared image processing technology wasutilized to identify the railway switches. By processing and recognizing two infraredimages of a switch, right direction of a switch could be decided.According to the actual characteristic of a railway switch, pre-process its image bymeans of gray processing, median filtering and histogram equalizing. Afterwards, byartificially segmenting the threshold value combining with the histograms, the image ofthe switch was binarized. The method of edge extraction is apllied by finding thedistinguished features of the binarized image. In this extraction process, Canny operatorand Sobel operator were used firstly, but the extracted segments were not good enoughfor later recognition. As a result, a new method was apllied. Firstly, the central axis wasthinned in the bianarized image. Then PPHT Hough Transform based on OPENCV wasused to wipe off the wasted lines. Finally, a series of processing procedures was appliedto get a fine image of railway lines. In the last part of the experiment, Triangulation wasused to fix the current rail of a train in infrared image. Next, to recognize whether or notthere is a small gap to judge the switch and its direction. For the recognition of the gap,we counted the pixels in the image. If the pixels in one side of a rail is much less thanthose of another one, the switch is judged to be switched in this direction, and the trainwill also turn into this direction.On the basis of the calculated results, the experimental results are in accordancewith the practical ones, which means this method achieved the goal.