Application Of Image Processing In Detecting Defects Of Catenary Hanger

The dropper is an important part of the overhead contact system.It is particularly important to detect the defective dropper regularly for the defective dropper will have a significant impact on the railway safety operation.Initially,the thesis introduces the Overhead lines,dropper and the image processing technology in detail,meanwhile,an image acquisition system based on the industrial camera in high frequency and high power is built.Secondly,the corresponding detection method of the defective dropper is designed which is divided into three parts: the detection of the dropper,the detection of the defective dropper and the positioning of the defective dropper.The preprocessing of the hanging string is mainly to reduce the influence of noise and illumination by means of image processing.At the same time,the main body part of the hanging string is strengthened,and the algorithm of threshold segmentation,image filtering,image brightness correction and image running blurred is used to improve the quality of the hanging string graph effectively.In this paper,the hanging string is located by the way of the indirect positioning clamp,and the advanced depth learning method is used to model the hoisting string clip.Faster RCNN is used to locate the clamp quickly,and then the area of the hanging string is cut.The defect detection of the hanging string is to deal with the small image of the hanging string,which is cut out.This paper uses the image processing and the depth learning to deal with the detection of the suspending chord defect.For the simple defects of the broken string and the force,the image processing scheme is adopted,and the algorithm is judged by using the algorithm of Canny edge and combining the algorithm of finding straight lines and curves.The method of depth learning is used to classify the strands of the hanging string and the loosening of the bolt.The test statistics and analysis are carried out through a large number of hanging string images,and the defect detection rate of the hanging string is more than 95%.Finally,through the statistical analysis of the hanging string data on the real section,it is proved that the suspension string defect detection system is reasonable and can detect the suspension chord defect accurately and effectively.