Detection method of rail slab disease based on significance analysis

Abstract

The invention discloses a track board disease detection method based on saliency analysis, which mainly comprises: (1) pre-processing the video by illumination, and extracting the spatial saliency ofthe vehicle-mounted video by fusing color, brightness and direction features. (2) on the one hand, sparse sampling being done to track video during the monitoring period, comparing the current monitoring video frame with the corresponding sampled video frame, and selecting the video frame with large variation as the reliable frame. On the other hand, a track video frame with abrupt change characteristics relative to the upper and lower frames is selected as a reliable frame. (3) Rough spatio-temporal salience map is obtained by fusing spatial salience map and temporal salience map. Consideringbackground prior, center prior and spatial compactness, the improved spatio-temporal saliency map is obtained for rough saliency map. (4) Significant weights of track diseases are obtained accordingto the saliency map, and the semi-supervised classification method based on saliency weighted model is used to identify and classify track diseases of high-speed railway. The invention can detect thetrack disease in real time.

Description

technical field [0001] The invention relates to the fields of computer vision and image processing, in particular to a method for detecting track slab defects based on significance analysis. Background technique [0002] The smooth passage of high-speed trains depends on high-quality high-speed railway tracks. Therefore, the quality of tracks is directly related to the safety of operations. However, due to exposure to the outdoors and the impact of various environments, high-speed railway track diseases occur frequently. Moreover, high-speed railways have only been put into operation on a large scale in recent decades, so the relevant disease detection work is relatively lagging behind, and the disease detection work in many places still needs to use expensive manual detection methods. In recent years, the continuous development of intelligent detection systems has made it possible to detect intelligent track defects. China has developed the GJ-6 track detection system usi...

AI Technical Summary

Problems solved by technology

[0004] (1) Manual inspection and traditional detection system can not deal with the track disease characteristics in a timely manner, and the results obtained at high cost cannot show the disease results well

[0005] (2) The existing detection method to process video data is to check and process each frame of video frame. This method not only has relatively high complexity, but also takes time to run. It is difficult to apply to occasions with high real-time requirements, and it cannot be well Highlight key disease areas

Images