Multi-moving target tracking method based on three-dimensional layered graph model

Abstract

The invention discloses a multi-moving target tracking method based on a three-dimensional layered graph model, and the method comprises the steps: S1, analyzing and detecting the appearance characteristics and moving characteristics of a target and the three-dimensional spatial structure characteristics among the characteristics, and building the target three-dimensional layered graph model of each target region; S2, establishing a detection three-dimensional layered graph model of a detection target in the detection area of the current frame; S3, establishing a prediction three-dimensional layered graph model of the tracking target in the prediction area of the current frame; S4, calculating the matching degrees of nodes, edges and space structures between the detection target in the detection three-dimensional layered graph model and the prediction target in the prediction three-dimensional layered graph model; and S5, carrying out tracking matching on the target between the detection area and the prediction area according to a calculation result of the matching degree. The method can effectively realize multi-target real-time tracking, not only can be applied to most Kinect monitoring scenes, but also can be popularized and applied to the fields of robot target recognition obstacle avoidance, intelligent transportation and the like, and has a relatively good application prospect.

Description

technical field [0001] The invention relates to the technical field of three-dimensional layered graph models, in particular to a multi-moving target tracking method based on a three-dimensional layered graph model. Background technique [0002] Multi-moving target tracking in a two-dimensional scene often cannot be tracked accurately due to mutual occlusion and loss of information. Therefore, the use of three-dimensional vision to achieve multi-target tracking in complex situations has attracted more and more attention. [0003] From the composition of the three-dimensional vision system, it is mainly divided into the following three types: monocular vision system, binocular or multi-purpose stereo vision system, three-dimensional depth RGB-D vision system, the monocular vision system only uses a two-dimensional camera, through three-dimensional calibration Obtain 3D visual information; binocular or multi-eye stereo vision system uses two or more cameras to jointly image an...

AI Technical Summary

Problems solved by technology

Through RGB-D data association and the optimization of the trajectory, the optimal association and tracking results of multiple human targets can be obtained, and the 3D point cloud information can be directly used to track according to the 3D features of multiple targets; however, kinect-based 3D visual analysis is currently mostly Concentrate on the 3D reconstruction of the scene, the navigation of the mobile robot, and the recognition and tracking. The multi-target tracking used in visual monitoring is in its infancy. Most of them use RGB and depth information to register 3D point clouds, and target recognition is performed on this basis. Tracking has high computational complexity and cannot be directly applied to the tracking of complex moving targets in video surveillance

Images