A detection method for large-scale difference parts of high-speed railway catenary

Abstract

The invention discloses a method for detecting large-scale difference parts of a high-speed railway catenary, which comprises the following steps: step 1: classifying data sets according to a clustering algorithm; step 2: forming a mapping relationship between N image categories and corresponding parts positions; Step 3: Input the data set into the convolutional neural network for classification training to obtain the classified data set; Step 4: Generate a rough positioning frame; Step 5: Extract the coarse positioning frame from the picture, and form fine positioning data according to the real positioning frame position Step 6: Input the fine positioning data set into the SSD positioning model to train the fine locator; Step 7: Input the image to be detected into the convolutional neural network to obtain the preliminary classification result; obtain the rough positioning frame of the parts ; Input the coarse positioning frame into the SSD positioning model obtained after step 6 training to obtain the fine positioning frame of the parts and complete the detection; the present invention is aimed at catenary parts with large-scale differences, with high positioning accuracy and short detection time .

Description

technical field [0001] The invention relates to the technical field of image recognition, in particular to a method for detecting large-scale difference parts of a high-speed railway catenary. Background technique [0002] The catenary support and suspension device is responsible for the important task of supporting the catenary, and the condition of the device affects the stability of the entire suspension system, thereby affecting the performance of the catenary wire; when a fault occurs, the catenary wire and the pantograph may not be in good contact , which affects the quality of the flow; among them, the catenary cable base has a harsh working environment, frequent and violent vibrations, sparsely populated along the road, and difficult maintenance, which is one of the weak links of the support suspension system; Mainly, the efficiency is low; at present, some researches have been done on the detection of catenary support and suspension devices based on image processing...

AI Technical Summary

Problems solved by technology

[0002] The catenary support and suspension device is responsible for the important task of supporting the catenary, and the condition of the device affects the stability of the entire suspension system, thereby affecting the performance of the catenary wire; when a fault occurs, the catenary wire and the pantograph may not be in good contact , which affects the quality of the flow; among them, the catenary cable base has a harsh working environment, frequent and violent vibrations, sparsely populated along the road, and difficult maintenance, which is one of the weak links of the support suspension system; Mainly, the efficiency is low; at present, some researches have been done on the detection of catenary support and suspension devices based on image processing. Traditional image processing is mostly for single component positioning. For example: Zhang Guinan proposed to use Harris corner detection and spectral clustering to realize the insulator anti-rotation matching; Han Ye proposed to use the local feature point matching between the catenary support suspension device image to be analyzed and the standard rotating binaural image to realize the positioning and extraction of rotating binaural ears; although deep learning can achieve multi-target positioning and The effect is improved but there are shortcomings; for example, Zhong Junping used three deep learning models to simultaneously locate 12 types of components of the high-speed rail catenary; but the experimental results show that the huge scale differences between different components seriously affect the detection results

Images