Dedect Recognition Of Catenary Rotating Binaural Based On Image Processing

The operation status of the catenary of the electrified railway is closely related to the driving safety.The high-speed trains obtain the electric energy through the catenary to ensure that the catenary is in normal working condition and it is essential for the train to receive electricity stably.The train runs under high-speed conditions for a long time.The vibration,shock and external environment are inevitable for the loss of the catenary support device.Therefore,it is very important for the defect detection and fault identification of the catenary support device parts.The traditional manual fixed-point detection and inspection vehicle routine inspection methods are difficult to meet the real-time and reliability requirements.With the development of related technologies such as high-definition imaging and image processing,inspection vehicles with non-contact detection methods have been put into production and application.However,the current detection systems generally have problems such as high missed detection rate and low recognition accuracy.With the large-scale construction of China's railways,the corresponding massive high definition photos of catenary support device also require the detection system to have higher recognition accuracy and detection speed.Therefore,it is necessary to study an effective automatic defect detection method for component characteristics.In response to the condition detection of pins that cannot adapt to traditional projection method in catenary rotating binaural,this paper proposed a method to detect the conditions of the pins based on CSGT(circularly symmetrical Gabor transform).The combination of the target detection algorithm SSD and the network of MobileNet and the rotation invariant LBP and HOG feature training SVM classifier are compared to realize the rotating binaural localization recognition.Using two methods based on convolution neural network and Hough transform to accurately position pins and split area of pins and utilizing CSGT to accomplish the feature extraction of edge Information.Finally,the BP neural network was used to judge and identify the condition of pins in catenary rotating binaural.The research and experiment showed that the method can accurately identify the positi on of the pins in catenary rotating binaural and possess high the recognition accuracy to the condition detection of pins.The method achieved the intelligent identification to the condition detection of pins in the complicated environment that bolts have random directions and various shapes.