Target tracking algorithm based on multiple vision in motion capture system

Abstract

The invention relates to a target tracking algorithm based on multiple vision in a motion capture system. The target tracking algorithm comprises the following specific steps: step1: carrying out binocular vision stereoscopic tracking, carrying out prediction by utilizing an extended Kalman, and picking out wrong candidate targets in a prediction area based on epipolar constraints; step2: carrying out multiple-view vision data fusion and processing a plurality of obtained binocular tracking data by utilizing a nearest neighbor data fusion algorithm based on variable threshold; and step 3: obtaining three-dimensional motion data of mark points. According to the target tracking algorithm based on the multiple vision in the motion capture system, the plurality of obtained binocular tracking data is processed by utilizing the nearest neighbor data fusion algorithm based on the variable threshold, thereby solving the problem of covering and loss of the mark points effectively.

Description

technical field [0001] The invention relates to the field of motion capture systems, in particular to a multi-vision based target tracking algorithm in the motion capture system. Background technique [0002] Marker-based optical motion capture system [1] Reconstructing the three-dimensional motion of the human body by capturing the marked points on the human body has been successfully applied in research fields such as film and television special effects, animation production, virtual reality, and simulation training. Existing capture systems usually employ feature-matching-based tracking methods [2-4] . Literature [2] uses different color blocks as marking points, and proposes a tracking method based on color blocks. Literature [3] uses a small number of colored ribbons as marker points, and uses human body structure constraint information as an auxiliary to track. Reference [4] uses the contour features of the marked points to track and achieves good tracking results,...

AI Technical Summary

Problems solved by technology

Reference [4] uses the contour features of the marked points to track and obtains good tracking results, but the marked points need to be specially made, which is not convenient for practical application. The tracking results of the above-mentioned color feature-based tracking algorithm are more accurate and can Solve the self-occlusion problem of marker points

However, because the flat color blocks are prone to deformation in 3D vision, there are differences in imaging hues between different cameras, resulting in inaccurate tracking, and when the number of marker points increases, the colors that can be clearly distinguished are limited, which greatly limits the their application

Images