Deep learning based track foreign matter detecting system

Abstract

The invention discloses a deep learning based track foreign matter detecting system. The system includes a vehicle image collecting device, an image transmission unit, a foreign matter detecting device, a machine learning device and an image database. The vehicle image collecting device is used for photographing and collecting images of a target track and above the track. The image transmission unit is connected with the vehicle image collection device and the foreign matter detecting device and is used for receiving collected images and transmitting the collected images to the foreign matterdetecting device. The machine learning device constructs an offline neural network model and utilizes images in the image database to train the neural network model. The foreign matter detecting device performs detection on track foreign matters based on the collected images and the constructed offline neural network model. The system provided by the invention is based on the neural network model.Compared with a traditional image recognition algorithm, better real time performance and accuracy are achieved and better adaptability and universality are achieved.

Description

technical field [0001] The invention relates to the field of track foreign object detection, in particular to a track foreign object detection system based on deep learning. Background technique [0002] Track foreign matter invasion refers to obstacles on the railway that affect the safety of trains, such as suspended objects dropped from bridges, tunnels, forests and trees, workers who are stranded on the track due to work mistakes, people and animals that are illegally on the track, and they leave garbage obstacles, etc. Due to the high speed of the train, it is difficult to guarantee the safety of the train only by relying on the traditional driver's vision and conventional detection system to identify foreign objects. Although the traditional railway disaster prevention alarm, fault diagnosis and detection theory and technology are relatively mature, foreign object alarm detection has always been a difficult point. With the development of new transportation technologi...

AI Technical Summary

Problems solved by technology

When cascade classifier training requires a large number of samples, a cascade classifier composed of several weak classifiers is obtained through a large number of sample training, which faces the problems of large image information to be processed and low operation efficiency, and foreign object invasion that can be detected Single and easy to misjudge

Images