Pantograph-catenary anomaly detection method based on template matching and neural network algorithm

Abstract

The invention provides a pantograph-catenary anomaly detection method based on template matching and a neural network algorithm, and the method comprises the following steps: obtaining related pantograph-catenary images which are gray images, and then classifying the images according to different scene sizes; using a template matching algorithm to intercept a bow net photo in each scene; resetting the sizes of the bow net photos, marking the bow net photos, and packaging all marked data into a data set which can be called; building a proper convolutional neural network; dividing the data set into a training set, a verification set and a test set, and then importing a training program to train the neural network; and obtaining a trained convolutional neural network and a corresponding template of each scene. In use, the image is grayed firstly, then the template matching algorithm is used for intercepting the bow net part of the image, then the convolutional neural network is used for carrying out state judgment on the image, and finally a judgment result is output.

Description

technical field [0001] The invention relates to the technical field of pantograph and catenary maintenance, in particular to a pantograph-catenary abnormality detection method based on template matching and neural network algorithm. Background technique [0002] Abnormal pantograph-catenary status is a common problem affecting power supply in traction power supply systems. These abnormalities sometimes only affect the use of pantographs, but sometimes cause serious parking accidents. Therefore, accurate identification of pantograph and catenary abnormalities is A very important requirement of the current traction power supply system. At present, in the actual application of railways, the image recognition technology for pantographs and catenary still stays on some traditional image recognition algorithms, such as template matching algorithm, edge recognition algorithm, etc., and even some places use artificial recognition The way to identify the captured image. At present,...

AI Technical Summary

Problems solved by technology

At present, for the research on pantograph-catenary anomaly detection, many studies are still committed to using traditional algorithms, "Powder-catenary detection and recognition algorithm based on high-definition image processing" ("Railway Rolling Stock", 2016,36(05):82-84 , Yang Luqiang, Han Tongxin) proposed a pantograph-catenary contact point detection algorithm based on the canny edge detection algorithm. This method is a typical image processing method based on traditional algorithms and is usually only applicable to specific and fixed scenes; "Image Recognition Based Catenary Information Processing System" ("Nanjing University", 2018, Hu Zunhao) proposed a contact line recognition algorithm based on robets operator and sobel operator. This algorithm can accurately identify the contact line value in the image, but it cannot The state of the contact line is judged, and some studies have begun to combine modern intelligent algorithms. "Research on Digital Image Processing and Recognition of Pantograph-catenary Operation State of High-speed EMU" ("Lanzhou Jiaotong University", 2020, Zhang Runtong) proposed a method based on Radon transform Pantograph-catenary operating state monitoring algorithm of convolutional neural network, which can identify the state of the pictures taken by the vehicle-mounted catenary operating state monitoring device (3C), but this method does not consider the situation of multiple scenarios, and does not make full use of it. Scalability of Convolutional Neural Networks

Images