A method and system for detecting foreign objects on rails from a space-based perspective based on weakly supervised learning

Abstract

The present invention provides a method and system for detecting foreign objects on a rail from an air-based perspective based on weakly supervised learning, which solves the problem that many types of foreign objects on a rail are difficult to identify using traditional detection methods. The system of the invention includes a space-based image acquisition module, an image preprocessing module, a two-level segmentation network module, a regional positioning module and an abnormal information output module. The method of the invention shoots the image of the ground rail area by the unmanned aerial vehicle, and after preprocessing the image, it is input into the two-stage segmentation network to extract the rough segmentation prediction map and the pixel sub-category prediction map of the rail area. The sub-category prediction map extracts the pixels actually belonging to the railway track in the map, and the invention trains a two-level segmentation network through a weakly supervised reverse learning algorithm, and reversely locates the abnormal area in the railway track according to the two prediction maps. The invention can effectively detect the foreign objects in the rails without marking the foreign objects themselves, and realize the identification and early warning of the intrusion of the foreign objects in the rails.

Description

technical field [0001] The invention belongs to the field of aviation surveillance, and in particular relates to a method and system for detecting foreign objects on rails from an air-based perspective based on weakly supervised learning. Background technique [0002] At present, railway construction has become an important part of the national development strategy. The convenient railway network provides a guarantee for national development and provides convenience for people's life. Therefore, how to effectively ensure railway safety is particularly important. In the daily operation of the railway, livestock and pedestrians may enter the track area and encroach on the track. At the same time, the leftovers left by railway workers on the track may also affect the normal running of the train. Therefore, in the daily maintenance of the railway, in order to ensure the normal and safe operation of the train, it is an important task to ensure that no foreign objects are invad...

AI Technical Summary

Problems solved by technology

[0006] Aiming at the current solution to the foreign object invasion on the rails, it is difficult to mark them and conduct training and testing through conventional detection models due to the uncertainty of foreign objects on the rails. This invention proposes a space-based detection method based on weakly supervised learning. The rail foreign object detection method and system from the perspective, based on the strong structural consistency and stability of the rail itself, learn to extract the normal area of ​​the rail, and then reversely locate the abnormal area to realize the detection of foreign objects on the rail

Images