Method and system for rail plug fault detection based on image alignment and detection network model

Abstract

The invention belongs to the technical field of computer fault detection, specifically a rail plug pin fault detection method and system based on image alignment and a detection network model. According to the invention, the method comprises the steps: recording a normal plug pin image as a reference image through a rail routing inspection vehicle; enabling a worker to obtain a rail condition image as a to-be-detected image in the routine rail routing inspection process through the latest rail conditions captured by the routing inspection vehicle, carrying out the comparison and analysis of the current to-be-detected image and the reference image, and detecting the position of a plug pin part with a problem on a rail, so as to timely replace the plug pin with the problem and guarantee the safety operation of the rail. The method specifically is based on the precise alignment of a depth network model with a high-speed rail image of a salient region and achieves the fault detection of the rail plug pin based on the depth network model and the environment change compensation. According to the invention, the method achieves the image alignment and environment change compensation based on the depth network model, and is more precise in recognition and higher in reliability.

Description

technical field [0001] The invention belongs to the technical field of computer fault detection, and in particular relates to a high-speed rail fault detection method and system. Background technique [0002] Railway is an important mode of transportation in our country. As the carrier of railway transportation, railway track is related to the safety of people's life and property. For a long time, the inspection of high-speed rail tracks has been carried out manually, which is easily affected by weather conditions, light, etc., and there are potential safety hazards for rail inspection personnel. High-speed rail track fault detection based on image processing has important research value. [0003] Since there are few high-speed rail track faults in practice, it is possible to judge whether there is a fault in the plug by taking pictures of the high-speed rail track at the same position in multiple months and comparing the changes in the plug. [0004] Plug pin fault detect...

AI Technical Summary

Problems solved by technology

There are three main difficulties in the alignment and detection of track images: First, the difference between the tracks is small, and direct matching is easy to cause wrong matching

The second is that the positioning information of the image is inaccurate. The rail inspection vehicle is equipped with a GPS system, but there are errors in GPS positioning

The third is that the high-speed rail track is greatly affected by the lighting factor. Due to the different shooting time and different lighting, the railroad track at the same position has a big difference in brightness

In the literature [6], the author proposed an object detection algorithm based on a deep network structure, but it did not include countermeasures for dealing with alignment and environmental changes in high-speed rail track images

Images