Traffic scene risk assessment method and system based on multi-branch convolutional neural network

Abstract

The invention provides a traffic scene risk assessment method and system based on a multi-branch convolutional neural network, and the method comprises the following steps: firstly extracting opticalflow information according to a video frame, enabling a video frame image to serve as space information, enabling an optical flow image to serve as time information, and inputting the space information and the time information into a space-time multi-branch convolutional neural network for learning and training; besides, adding a time shifting module and an attention module on the basis of the convolutional neural network: the time shifting module realizes information exchange of spatial and temporal features between adjacent frames without increasing the number of network parameters, and theattention module can learn a specially changed region in the scene, so that the risk estimation accuracy is improved; using a sparse time sampling strategy in the training and testing process, dividing a video into multiple segments, extracting video frames and inputting into a network, so that the calculation speed in practical application is increased while information redundancy between adjacent frames is avoided.

Description

technical field [0001] The present disclosure relates to the technical field of traffic scene risk assessment, in particular to a traffic scene risk assessment method and system based on a multi-branch convolutional neural network. Background technique [0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art. [0003] With the continuous development of the global economy, the number of cars and drivers is increasing year by year. According to the data released by the Ministry of Public Security, the number of cars in my country has exceeded 200 million, and the problem of road traffic safety has become increasingly prominent and has attracted widespread attention. . The latest statistics from the World Health Organization show that the number of deaths due to road traffic accidents in the world is about 1.25 million every year, and my country is the country with the most road t...

AI Technical Summary

Problems solved by technology

The inventors found that although the road monitoring camera has the advantages of fixed video background and appropriate angle, it can detect or predict dangerous driving behavior by using some traditional image processing methods or simple neural networks, but the detection results of the road monitoring camera can only be Provides a wide range of road risk assessments, cannot effectively assess road risks at local locations, and cannot effectively provide the assessment results to drivers in real time

Based on the above problems, some researchers have proposed a traffic scene risk estimation method for the video shot by the car camera. However, due to the problems of background motion, video jitter and shooting angle in the video shot by the car camera, the feature extraction model design process of the method is relatively difficult. Complicated, unable to obtain good evaluation accuracy, and the calculation time is relatively long

Images