Detection Of Catenary Insulator Crack And Locat Supportor Based On Image Processing

The good working condition of the catenary and its ancillary devices is the basic guarantee for the safe operation of high-speed trains.Because of its long working time,the catenary insulator and locat supportor are constantly affected by electrical shock and mechanical stress,if the insulator and locat supportor in the event of a failure,then the catenary equipment will be damaged,while the direct will lead to High-speed train stops or even cause casualties.At present,based on image processing technology insulator working state detection has been studied,which greatly enhance the efficiency of equipment inspection,saving a lot of manpower and resources.However,the previous detection method is basically based on the study of insulator missing or inclusions.The detection of insulator fracture is less and the detection of the locat supportor is almost blank.Therefore,it is necessary to use the image processing machine learning efficient method to detect insulators and the locat supportor.In the study of the previous catenary and ancillary device detection methods,the identification of insulators in the power system and contrast face recognition method,this paper use the image data collected by the catenary integrated patrol car as the original image of the sample,and then the texture feature LBP of the insulator used to training classifier by the machine learning to extract the insulators in the catenary images and then analyzed the insulater crack.At the same time,four kinds of feature matching methods based on ASIFT,SURF,ORB and SURF are used to detect the locat supportor.Firstly,the pre-processing of the original image of the insulator is carried out.The positive and negative samples are based on a large number of insulator and non-insulator images.Then,the LBP and HOG characteristics of the insulator are extracted.The method of machine learning is used to extract the target insulator in catenary images.The features of LBP and HOG are given to Opencv by using the Adaboost algorithm to train the classifier.Then,the classifier model is used to identify the insulator in the image.The model combined LBP and Adaboost are the most efficient.Finally,this model is used to extract the insulators precisely for a large number of catenary images,using the threshold method to binarize the intercepted insulators,divide the insulators and crack characteristics.The geometrical characteristics of the insulator crack are calculated by using the method of finding the area of the connected domain,so as to realize the crack detection.In the recognition of the locat supportor,a variety of feature matching algorithms are used to locate the position,the SURF operator proved to be the best performer.The experiment is based on the OpenCV2.4.13 library and the software VS2013 programming.Through the experimental test of a large number of catenary images,the conclusion is drawn that the validity of LBP and Adaboost model in insulator detection,the accuracy of insulator crack analysis and the reliability of SURF operator on locat supportor detection.