Railway shunting signal lamp detection method and system based on binary convolutional neural network

Abstract

The invention provides a railway shunting signal lamp detection method and system based on a binary convolutional neural network, and the method comprises the following steps: S1, data set preparation: collecting a train head front video obtained during the running of a train, cutting the video into pictures with a set size, and manually screening pictures containing a target, thereby obtaining atarget picture; dividing the target picture into a training set and a test set according to a set proportion of a blue lamp and a white lamp; S2, network construction: constructing a binary convolutional neural network framework by utilizing the training set; S3, network training: training a binarized convolutional neural network by using a binarizing method through the test set to obtain a targetnetwork; and S4, performing network operation, and performing real-time target detection on the railway shunting signal lamp by using the target network. Binaryzation is carried out on the neural network weight to accelerate neural network operation and reduce memory consumption of the weight. The invention is high in accuracy, high in reliability and capable of accurately detecting the signal lamp in front of the train.

Description

technical field [0001] The invention relates to the field of image detection, in particular to a method and system for detecting railway shunting signal lights based on a binary convolutional neural network. Background technique [0002] my country's railways are developing rapidly, and the annual train freight volume continues to rise. Under the environment of rapid development of railways, railway shunting accidents occur frequently, causing heavy life, economic loss and serious social impact. The main reason is the manual misjudgment of the status of the signal lights by the flight attendants. [0003] Convolutional neural networks have greatly promoted the development of tasks in the field of computer vision. Convolutional neural networks have demonstrated their reliable performance in the field of object recognition and detection, and have been used in real-world applications. At the same time, using a large number of image resources to train neural network models is ...

AI Technical Summary

Problems solved by technology

Recognition systems based on full-precision CNNs require large amounts of memory and computing resources, but they are usually not suitable for smaller devices such as mobile phones and embedded electronics

Images