Inspection method of railway traffic catenary based on full convolutional network of attention mechanism

Abstract

The invention provides a railway traffic catenary inspection method based on an attention mechanism full convolution network, which belongs to the field of aviation surveillance. The method of the present invention includes: the unmanned aerial vehicle inspects the railway, takes the video and returns it to the ground station server; the ground station server performs frame extraction and preprocessing on the video, inputs the trained attention mechanism full convolutional neural network, and obtains the railway in each frame image The pixel-level distribution probability map of the catenary, and then use the LSD straight line extraction algorithm to extract the straight line segments belonging to the railway traffic catenary, connect the straight line segments into curves and perform curve fitting to detect and alarm potential safety hazards. The invention solves the problem that it is difficult to distinguish the catenary in the inspection of the railway catenary on the air-based platform, reduces the operation and maintenance cost of the traditional patrol inspection, and can timely check the problem of the catenary line to improve the safety of the railway operation. It is of great significance for the daily operation and maintenance and safety warning of railways.

Description

technical field [0001] The invention belongs to the field of aviation monitoring, and relates to a method for inspecting a railway traffic catenary based on an attention mechanism full convolution network, which is used for patrolling a railway traffic catenary line based on an air-based platform. Background technique [0002] During the daily operation and maintenance of the railway, it is necessary to regularly check the status of the railway line to check whether there are foreign objects attached, whether there are abnormal conditions such as hanging and falling off. Early detection of these abnormal situations can effectively guarantee railway operation and prevent unnecessary casualties and property losses. [0003] Traditional staff inspections not only waste a lot of manpower, but also are slow and inefficient; roadbed inspections require rail cars to inspect along the line, the inspection time is limited, and may affect normal railway operations. [0004] The space...

AI Technical Summary

Problems solved by technology

[0006] The present invention aims at the problem that the catenary in the image collected by the drone is very thin in the current space-based catenary inspection, and it is difficult to distinguish the catenary and the background in the collected image in the face of the complex environment on the ground. A Patrol Inspection Method for Railway Transportation Catenary System Based on Full Convolutional Network of Attention Mechanism

Images