A method and system for identifying railway foreign objects based on machine learning

Abstract

The application provides a machine learning-based method and system for identifying foreign objects on a railway, the method comprising the following steps: during the running of the train, collecting a monitoring video of the track in front of the train, and obtaining multi-frame monitoring images in the monitoring video in real time; Establish a dangerous area foreign object detection window in the image, and extract the track dangerous area detection image whose monitoring image is located in the dangerous area foreign object detection window; preprocess the track dangerous area detection image; input the preprocessed track dangerous area detection image to In the pre-established foreign object recognition and classification model, the foreign object recognition and classification model recognizes whether there is a foreign object in the detection image of the dangerous area of ​​the track. If there is a foreign object, an alarm signal is sent, otherwise, continue to identify other detection images of the dangerous area of ​​the track. The application has high detection accuracy for railway foreign matter, saves manpower, automatically alarms, has low cost, and is less affected by the environment.

Description

Technical field [0001] The present application relates to the railway foreign matter intrusion detection technologies, and in particular to a method and system for railroad foreign object detection based on machine learning. Background technique [0002] Wide distribution area of ​​high-speed railway, long distance, complicated geological conditions, natural disasters, serious, while the highway has day and night operations, closed, allowing only train travel and so on. Due to the long distance, wide distribution area, around the clock, closed, etc., when the high-speed railway landslides, subsidence and other geological natural disasters, can not be published in time road traffic fault warning, easily lead to serious traffic accidents and a chain reaction. In order to reduce high-speed rail accidents, reduce casualties and property losses, the urgent need to address natural disaster warning along the high-speed railway and special traffic accidents to the police. [0003] Such a...

AI Technical Summary

Problems solved by technology

[0003] Rolling stones, pedestrians or animals and other foreign objects invade the railway boundary, which is sudden, irregular and unpredictable, frequently causing railway traffic accidents and seriously threatening the safety of people's lives. Traditional track detection mainly relies on manpower. Setting up a large number of inspectors to check the intrusion of the track, but it consumes manpower and financial resources, and the response to emergency accidents is not fast enough

[0004] In addition, the video surveillance system is used for railway safety monitoring to monitor whether there are foreign objects intruding into the railway boundary. However, the current video surveillance system requires special personnel to monitor, and the workload of the monitoring personnel is heavy. When they are tired, omissions are prone to occur, causing danger. Hidden danger

The existing method of using a laser curtain wall for foreign object monitoring, this method detects intruding foreign objects by installing multiple two-dimensional laser sensors to form a laser curtain wall, this method has a fast detection speed and high sensitivity, but the installation is relatively complicated and is greatly affected by the environment. Only a few sections can be detected, and the cost is high

Images