Research On Railway Guardrail Online Detection System With Complex Background

With the development of high-speed railway industry and the continuous extension of the railway industry in our country,in order to ensure the safety of railway transportation and detect the surrounding environment of railway becomes more and more important.But existing comprehensive inspection train was just detection the tracks without railway guardrail which is detected by workers.So,the detection efficiency is low and labor intensity.Moreover,if the railway guardrail were damaged,the small animals may lodge in the railway line.It will cause significant security risks.Therefore,the detection of railway guardrail is particularly important.For the complex background,this paper extract railway guardrail from different aspects.This paper proposes a novel algorithm for extract railway guardrail,it provided theoretical supports for the railway guardrail online detection system.The following research works have been done in this paper:Firstly,Proposed the system frame of the railway guardrail online detection system,the system main include image capture system,image process system and wireless transmission system.Due to the requirement of system and determine the main hardware model.Secondly,described the principle of digital image processing methods and used it to pre-treatment for original railway guardrail image.In order to extract railway guardrail easily filter the noise.Thirdly,according to the characteristics of railway guardrail,using edge detection method,respectively using Sobel edge detection,Canny edge detection and edge detection method based on active contour model to edge detection the railway guardrail image.Digital image processing usually converted RGB color image into gray image,it lost a lot of color image information.So using color feature extraction to extraction the railway guardrail,and achieved certain results.But the railway guardrail the specification and the color is not uniform,so that it not suitable for the whole detection system.According to the simulation results,edge detection method handle too detailed and did not feature selection so that complex background was detected,it will caused major problems on extraction grid.So that just using simple edge detection it will not extraction the all of railway guardrail.And base on the basis edge feature extraction and color feature extraction results show that railway guardrail has symmetry,so according to the symmetry proposed the method to extract railway guardrail.Finally,proposing a novel railway guardrail extraction algorithm.The algorithm was mainly divided into three parts.Part one,through the Harris corner detection to detection all the obvious features of the image,and then clustering the corner point with K-means,and find the grid intersection,and using the A-score to ranking the cluster.Then the highest rank was the grid intersection point.Connecting these corner points,propose the initial model and edge vector of grid.Part two,According to the Markov random field,belief propagation and mean shift proposed mean shift belief propagation(MSBP)algorithm for grid expansion,finding the all grid.Part three,verify the grid growth by MSBP,and determine whether more gird texture features.According to the experimental results show that the proposed grid extraction algorithm can effectively extract the railway guardrail from complex background.