Research On Detection And Identification Of High-speed Rail Arch Line Based On Hough Transform

The touch line is an important part of the contact net,when the touch line malfunction occurs,the normal operation of contact net will be affected,even the high-speed train will be interrupted.Therefore,the research of how to improve the detection and recognition of touch line veracity is significant important to high-speed rail.Firstly,the paper introduces the research status at home and abroad and the difficulties in detection and identification.It also introduces the existing method of detecting and identifying the touch line.Secondly,it designs the corresponding method for detecting and detecting the touch line.It is divided into: infrared image preprocessing,The extraction of the pantograph and the detection of the archwire identify three parts.First,since the infrared camera located on the roof collects the arch image all day,the original image contains a large amount of noise and uneven illumination.Therefore,the image must be preprocessed first.The preprocessing steps mainly include adaptive median filtering to denoise the image.For the problem of uneven illumination,the morphological method is used to remove the image.Secondly,in the extraction part of the pantograph,after observing a large number of original infrared images collected on the spot,it is found that the edge features of the pantograph are obvious and hardly deformed when the train is running.Therefore,the edge detection method can be used to extract the upper edge of the pantograph.However,if there is a lot of lateral interference in a complex background,such as above the pantograph,only the edge detection to extract the pantograph will greatly reduce the accuracy.Therefore,this paper selects the template matching algorithm based on normalized cross-correlation(NCC)to extract the pantograph.Although the complexity of this method is larger than that of edge detection,it has strong anti-interference ability,good robustness to illumination changes and geometric distortion,and the extraction accuracy is significantly improved.Finally,for the problem of relatively large complexity,two directions are proposed to solve.Thirdly,in the detection and recognition part of the touch line,this paper firstly uses the recursive OTSU threshold segmentation and the connected domain denoising based on the seed filling method to achieve accurate extraction of foreground images.Then,the binary image is refined to improve the accuracy of the intersection of the touch line and the pantograph.Finally,the hough change and the straight line clustering of the straight line to be tested are successfully detected by the angle and brightness characteristics.Identification.Finally,the proposed method is tested in different backgrounds,and the continuous image of 6825 frames collected in the field is used for testing,and compared with other methods,it proves that the design method is reasonable and accurate.The detection identifies the touch line.