Corbelling Detection Of The Catenary And The Serial Number Recognition Based On Image Processing

The rapid development of China’s high-speed railway puts forward increasingly high quality requirements for operator. Traditional catenary detection methods are inefficient and have great error, which gradually can’t meet the demand. At the same time, the passive detection method, based on intelligent video surveillance technology begins to receive more and more attention because it can avoid being involved into existing systems, as well as little impact on traffic, highly speed, being in line with the human perception. Today’s intelligent video surveillance technology, help us to understand the meaning of the image content, explain the objective scene information through locating and identifying the image of the target scene.Use this method to detect high-speed rail power supply safety monitoring system, you can gradually replace the existing manual inspection method, making the detection monitoring systems more automated and intelligent. So how to efficiently identify the catenary inspection images with complex background is making great significance for inspection technology of high-speed rail.This paper relies on the sub-catenary safety inspection devices of the high-speed railway power supply safety monitoring systems, and is based on information about the railway scene sequence of video images which is get by the devices. We make full use of image processing techniques, attend to the research of the detection of catenary pillars and catenary pole number, propose solutions and verify the accuracy of the method by experiments.In the detection of Catenary pillar, we take advantage of the rail feature to strike Vanishing Point, provide location information and divide the contact line regional and rail, eliminating interference. We use motion characteristics to compute optical flow, get the range of the pillars motion and get rid of complex background interference. Then combine with Lucas-Kanade optical flow field and the probability Hough transform for line detection catenary pillars to position. Finally, according to catenary pillars build the model geometry, complete catenary pillars of model reconstruction, lead to the purpose of the pillars detection. In identification of Catenary pole number, locate the Catenary pole number according to the pillar location information and histogram features. Then enhance image information via adaptive binarization, thin algorithm to extract frame structure, fine line detection region segmentation accurate rod number, get accurate rod number regions which are divided finely by line detection. Finally, extract a single character exactly based on statistical value of image gray, and character recognition, to get pole number of digital information.Methods presented in this paper are validated by experiment with the data of existing railway real scene images. Complete the test of existing data sets and analyze the experimental results. Through the experimental data, results show that:the method put forward in this paper, in the actual scene of the running train, can accurately detect the catenary pillar and identify the corresponding contact net rod, and has high accuracy. What’s more, the algorithm of fast processing speed, can meet the requirements of real-time processing, meanwhile, it has certain engineering application value.