The Design And Implementation Of Catenary Fault Detection System

Catenary is an important part of electrified railway power supply system.If the catenary fails,it will directly affect the electrification of the railway transportation,in the railway transportation accident,the contact network fault accounts for more than sixty percent,so to ensure the normal work of the catenary is the premise to ensure the safety of railway transportation.Manual inspection waste a lot of manpower,material resources and time,and the effect of inspection can not keep up with the demand of railway development.Therefore,the research of catenary fault detection is of great significance and practical value.The method an image similarity comparison of the nearest neighbor difference and a false positive filtering scheme based on Convolutional Neural Networks were present,the system is divided into two modules: catenary pole number identification and catenary fault detection.The two modules are as follows:First,Single Shot Multi-Box Detector is used to identify the string number character,and the identified letters or numbers are added to the character queue in order to connect the characters,to realize the contact network bar identification module,and to locate the specific location information of the contact network.Secondly,due to the cyclic characteristics of the insulator components,the detection method of the feature class will produce large error.Therefore,according to the classification information of the contact network components,different types of key components are detected by different methods.The Oriented FAST and Rotated BRIEF feature point matching algorithm and the image difference method based on near neighbor idea are used to detect the key components.The Label Consistent K-Singular Value Decomposition algorithm and the image difference method are used to detect the insulator components.The fault region is added to the fault output chain list.In addition,the detected fault areas are filtered to reduce the false positive rate.The image is processed by corrosion and expansion,and the fault areas such as slender and few pixels can be filtered out.For the larger fault area,the filtering scheme based on the similarity of convolution neural network is used to.Finally,the unfiltered area of the barrier area which is still unfiltered in the output chain is exported and the fault detection module is completed.The catenary fault detection system designed and implemented in this paper has been applied to some railway administrations.The results show that the manual inspection and the previous system show obvious advantages.