Road traffic accident judging method and device

Abstract

The invention discloses a road traffic accident judging method and device. According to the invention, lane lines are extracted from video images through obtaining the video images of a monitored road section; performing multi-target detection tracking on vehicles on each lane and determining driving tracks of the vehicles; according to relative positions of the driving tracks of the vehicles and the lane lines, judging whether the vehicles have lane change or not and determining lane change when the driving tracks of the vehicles deviate from the range of the current lanes; judging whether the vehicles have sudden stop or not according to coordinate differential values of vehicle feature points in video images of adjacent frame numbers and determining the sudden stop when the coordinate differential values are smaller than a preset threshold value; and determining traffic accidents in the monitored road section when detecting that multiple vehicles in one lane in the monitored road section have lane change and / or sudden stop. The method and device provided by the invention improves accident judging accuracy and can handle extreme traffic accidents with huge deformation of vehicles and the like well.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a method and device for judging road traffic accidents. Background technique [0002] The modern intelligent traffic video monitoring system has the ability to obtain rich information including lane congestion and vehicle violations more accurately and comprehensively, and can monitor various traffic conditions required by multiple lanes in real time. It has very good application and research value. [0003] For the identification of road traffic accidents, the current intelligent traffic video surveillance system mainly focuses on the feature extraction of the foreground targets of existing traffic events, and performs vehicle size segmentation for various types of vehicles including cars, buses and trucks. The traffic incidents are judged and identified by using the closed value of the area and the closed value of the width-to-length ratio of various vehicles in the de...

AI Technical Summary

Problems solved by technology

Due to the need to train multiple images and use a classifier for real-time classification in this existing traffic accident discrimination method, the information processing speed is slow, and it lacks universal application for special traffic events, such as large deformation of vehicles, etc. sex

Images