Research Of The High-Speed Railway Catenary Suspension Object Detection Based On Deep Leaning Image Processing

The rapid development of high-speed railways and the demand for operational quality have placed higher demands on the safe operation of power supply equipment for traction power supply systems.The catenary is a special type of transmission line that is provided along the railway line to supply power to the electric locomotive.Because the catenary has no backup,and it needs to withstand the dynamic take-off of the pantograph of the high-speed train,its state directly affects the traffic.Since 2016,the 4C devices that have been put into operation in railways have used image matching algorithms such as template matching in the intelligent image recognition technology.Although the efficiency of the traditional manual inspection has been improved,there is a poor generality for different lines.,Accuracy is difficult to improve,the number of automatic identification algorithms is limited,and computational efficiency is low.Further research is needed to improve the above problems,so as to make the 4C device more valuable.Therefore,it is necessary to study the new intelligent image recognition algorithm for high-speed railway contact net suspension state.This paper firstly analyzes the requirements of intelligent image recognition for the detection and monitoring device of the catenary suspension state,and proposes the basic objectives of the image algorithm based on the types of high-speed rail contact nets that have been built in China and the typical defects of common support devices.Then based on HALCON,a positioning algorithm based on template-matching flat wrist arm U-type hoop and a traditional wrist-arm U-type hoop pin missing detection algorithm based on connected-domain analysis and local dynamic threshold were implemented..The algorithm can be applied to the detection and monitoring device of the catenary suspension state.Although it has achieved certain results,it can not fully meet the requirements of the smart image recognition of the catenary suspension state detection and monitoring device.In order to circumvent the traditional image detection methods,Poor performance,low development efficiency,etc.,give full play to the value of intelligent image recognition,improve the analysis efficiency of detection data,and propose a positioning algorithm based on deep learning for high-speed rail contact net support device components.In deep-learning based on high-speed railway contact network support device component positioning algorithm,based on the Caffe deep learning framework,Faster R-CNN target detection algorithm based on VGG16 basic network is selected to produce a sufficient number of image data sets and complete the convolutional nerve Network training,training convolutional neural network model.Finally,using the trained model to locate the supporting device components of the acquired high-speed rail contact network support device images of multiple actual lines,and comparing and analyzing the experimental results with the traditional image algorithm,the high-speed rail contact network support device based on deep learning was verified.The feasibility and effectiveness of the component positioning algorithm can be extended to achieve the positioning of other components,thereby improving the efficiency of detecting image data.