Intelligent Detection Of Catenary Steady Clamp Based On OpenCV

As the core of high-speed railway overhead catenary, the state of catenary support and suspension devices directly determines the stability of whole suspension system, and effects quality of current collection of pantograph and catenary. Steady clamp is the key component to connect support devices and contact line, and withstands electrical shock and mechanical stress, whose nut is prone to lose or shed. Once faults occur, the safety and reliability of traction power supply will be affected. Currently, mainly two methods to detect steady clamp state are manual inspection and viewing photos, with low efficiency and accuracy. And, the catenary suspension state detecting and monitoring devices in practical applications is focused on insulator fault detection and steady arm slope detection. However, there is no practical application for the detection of the steady clamp bad state. Therefore, it is necessary to study the intelligent algorithms to detect steady clamp image.This paper analyzed the successful application of image processing in insulator fault identification, geometric parameters of catenary and steady slope detection, and proposed target recognition and state detection algorithm for steady clamp image based on image processing. In addition, completed system design and user interface development, and the actual line image results verified the practicality and versatility of the method.Firstly, preprocess image to realize de-noising, enhancement and segmentation. Then, achieve steady clamp recognition and extraction based on structural feature detection and machine learning. Finally, analysis and process the steady clamp local region image, and realize defects identification and warning.In steady clamp recognition, design support structure line detection algorithm based on Hough transform and morphological processing. Then, detect structure feature and extract candidate region in line endpoint region. Finally, according to image HOG features and SVM classifier,realize steady clamp identification and location in the candidate area.In steady clamp state detection, preliminary position and precisely extract nut regions by steady arm lower edge extraction and MSER algorithm. Secondly, analyzed nut state based on the combined results of aspect ratio, rectangle degree and pixel statistical value. Lastly, warn the defect steady clamp.The steady clamp recognition and state detection algorithm is realized by programming based on Microsoft Visual Studio and OpenCV, further developing user interface. Finally, this paper verified the reliability and feasibility of steady clamp intelligent detection algorithm and system, through testing inspection images of catenary suspension state detecting and monitoring devices under different conditions of several railway lines.