A video driver behavior identification method based on a multi-task space-time convolutional neural network

Abstract

The invention provides a video driver behavior identification method based on a multi-task space-time convolutional neural network. A multi-task learning strategy is introduced into a training processof a space-time convolutional neural network and is applied to driver behavior identification in a monitoring video. Auxiliary driver positioning and optical flow estimation tasks are implicitly embedded into the video classification tasks, so that the convolutional neural network model is promoted to learn richer driver local space and motion time characteristics, and the driver behavior recognition accuracy is improved. Compared with an existing driver identification method, the multitask space-time convolutional neural network architecture designed by the invention combines interframe information, is high in generalization and identification accuracy, can be used for real-time driver behavior identification under a monitoring video, and has an important application value in the field of traffic safety.

Description

technical field [0001] The invention belongs to the field of image processing and pattern recognition, and relates to a video driver behavior recognition method based on a multi-task spatio-temporal convolutional neural network. Background technique [0002] According to the official report of the World Health Organization, 1.25 million people die in traffic accidents every year in the world. As one of the most frequent accidents, the occurrence of serious traffic accidents is usually attributed to drivers' illegal driving, car failure, bad weather conditions, etc. Among them, more than 80% of traffic accidents are related to drivers' illegal driving. Some bad driving behaviors, such as taking hands off the steering wheel, making phone calls, looking down at mobile phones, smoking, etc., distract the driver's attention and pose a certain degree of safety hazard. Therefore, driver behavior monitoring technology has important research significance for road safety and intellig...

AI Technical Summary

Problems solved by technology

So it is difficult to artificially design general features suitable for all scenarios

[0005] (2) Although the traditional convolutional neural network method can be applied to static image recognition, it cannot make good use of the dynamic information between videos

The two-stream convolutional neural network can extract the dynamic information between frames through the dense optical flow map for classification, but because the optical flow map needs to be pre-calculated, it cannot realize real-time recognition of driver behavior

Images