Method and device for acquiring driving data of vehicles in multi-camera videos

Abstract

The invention discloses a method and a device for acquiring driving data of vehicles in multi-camera videos. According to the method, multi-camera road traffic observation videos are stitched into a whole, a uniform coordinate system is established, driving data of the vehicles are extracted, and the problem that the multi-camera videos are difficult to process during traffic information acquisition is solved. The method comprises steps as follows: the splicing point of each camera video is selected, coordinates of an original image at each stitching point are determined, coordinates of a road at each stitching point are determined, coordinates of an image stitched at the stitching points are determined, an image transformation matrix is calculated, video image stitching are performed, and vehicle data are extracted. The method is simple and convenient in design, high in adaptability and efficiency and easy in program realization, multi-camera traffic videos can be effectively treated, complete extraction and coordination of the driving data of the vehicles are realized, and following work including driving behavior study, traffic condition analysis, traffic state prediction, traffic management and the like is greatly facilitated.

Description

technical field [0001] The present invention relates to the field of traffic observation, in particular to a method and device for extracting vehicle driving data from road traffic videos, which can realize the splicing of road traffic videos of multiple cameras into a unified whole, and continuously extract vehicle trajectories, speeds, acceleration data. Background technique [0002] Vehicle driving data is an important basis for driving behavior research, traffic status analysis, traffic state prediction, traffic management, etc. With the advancement of video observation technology, video observation has become an important way to collect vehicle driving data, and the effective extraction of vehicle driving data from video is the key link. When the observation range is large, it is often necessary to set up multiple cameras to shoot videos together. The same vehicle may appear successively in the videos of multiple camera positions. If the relevance of the videos of eac...

AI Technical Summary

Problems solved by technology

In this method, vehicle comparison is difficult and inefficient, and the video coordinate system of each camera position is not uniform, and coordinate conversion is difficult, so it is often difficult to obtain relatively complete vehicle driving data

Images