Open area target detection and tracking method based on binocular vision sparse point matching

Abstract

The invention relates to an open area target detection and tracking method based on binocular vision sparse point matching. The open area target detection and tracking method based on the binocular vision sparse point matching comprises the following steps. Calibration of video cameras is achieved by utilizing a Zhang checkerboard calibration method, and a transformational relation between a video camera coordinate system and a user-defined world coordinate system is established. Afterwards, related algorithm of the method is executed, wherein synchronous images of two video cameras are input. The images are first corrected to an extreme line parallel state, and then sparse characteristic point matching is achieved by utilizing extreme line constraint. Matching points are mapped to the world coordinate system and projected to a ground plane. Discrete point clustering is achieved by means of human body type information and a golden section ratio. Finally, robust detection and tracking of pedestrian targets in a monitoring area are achieved with the clustering as a unit and by combining joint probability data association and target color information. According to the open area target detection and tracking method based on the binocular vision sparse point matching, the operating amount is enabled to become small, and the target detection and the tracking are enabled to be accurate.

Description

technical field [0001] The invention relates to a target detection and tracking method in the technical field of computer vision, in particular to an open area target detection and tracking method based on binocular vision sparse point matching. Background technique [0002] Moving target detection and tracking has always been an important research direction in the field of computer vision, and it is also the basis for target behavior analysis and intelligent video event detection in the field of video surveillance. However, the existing target detection and tracking algorithms are far from meeting the needs of intelligent analysis of video surveillance, and it is particularly important to improve this aspect of the work. [0003] Moving target detection methods mainly include optical flow method, frame difference method and background difference method. The optical flow method has high computational complexity and is usually not used; the frame difference method mainly use...

AI Technical Summary

Problems solved by technology

Philip.K et al. completed pedestrian detection based on the binocular dense disparity map based on human biological information (according to the golden section ratio). Mis-matching occurs; Lenz.P et al. proposed a sparse-point three-dimensional optical flow target detection and tracking method based on binocular stereo vision, which requires the calculation of a large number of continuous multi-frame optical flow data of feature points, and completes the target according to the speed vector difference of each point Segmentation is not suitable for targets with poor texture or non-rigid objects; Cai.L et al. proposed a target detection and tracking method based on sparse matching points, mainly based on the MeanShift algorithm to cluster discrete points to complete target detection, and Kalman filter for Target tracking has higher accuracy than dense matching, and overcomes the problem of sudden illumination changes. However, this method is prone to fall into local extremum and cause clustering failure

Images