Pedestrian target movement track acquisition method and system based on multiple cameras

Abstract

The invention discloses a pedestrian target movement track acquisition method and system based on multiple cameras. The method comprises the following steps: acquiring a video through calibrated cameras, and preprocessing the video to obtain a training sample set and a test sample set; constructing, training and testing a deep convolutional neural network model; matching the feature vector and theimage coordinate of the pedestrian target output by the model, converting the image coordinate into a space coordinate, and connecting the space coordinate of the pedestrian target in series according to the time sequence list to obtain a moving track of the pedestrian target under the single camera; in combination with calibration external parameters, realizing pedestrian target cross-camera association through feature matching and overlapping region position information, and connecting spatial coordinates of multiple pedestrian targets under multiple cameras in series to form a multi-pedestrian target moving track. According to the invention, the two-dimensional image coordinates of the pedestrian target are converted into the space coordinates in the actual scene, so that the positionand track of the pedestrian target can be accurately obtained; the accuracy of pedestrian target recognition and track acquisition is improved by using multi-angle video fusion analysis.

Description

technical field [0001] The present invention relates to the field of video monitoring and analysis, in particular to a multi-camera-based method and system for acquiring moving tracks of pedestrian targets. Background technique [0002] At present, video surveillance technology is very popular and has been applied to security prevention and control, work inspection, violation monitoring, etc. Common application scenarios include streets, shopping malls, schools, retail stores, offices, Internet cafes, etc. [0003] With the continuous upgrading of computer hardware equipment, the computing speed has been greatly improved, the price of video surveillance cameras is getting lower and lower, and the quality of video images is getting higher and higher. Algorithms such as target detection and recognition are more applicable in the field of video surveillance. This application can be extremely Minimize the cost of manual screening, shorten the target monitoring time, and greatly...

AI Technical Summary

Problems solved by technology

[0004] The present invention provides a multi-camera-based acquisition method and system for pedestrian target moving trajectories in real and complex scenes, which can solve the current problems of low efficiency in calculation and analysis of pedestrian target tracking and the inability to locate the actual position of pedestrian targets

Images