Three-dimensional Motion Capture Method For Particle Feature Objects Based On Binocular Vision

As an important part of computer vision,3D motion capture technology has broad development prospects in many fields,such as film special effects,virtual reality and intelligent control.Due to the high cost of motion capture equipment and the difficulty in achieving accurate rates,the application of this technology are limited to a degree.In order to improve the accuracy of the data,a 3D motion capture method for particle feature objects based on binocular vision is proposed in this paper.The main works are as follows:Firstly,the related backgrounds and techniques are expounded,such as the overview of binocular vision,the classification and steps of camera calibration and the introduction of motion capture methods.Then,the framework of motion capture method is proposed based on the theoretical basis.Finally,the spatial 3D reconstruction and temporal series reconstruction are used to achieve motion capture.In spatial 3D reconstruction,the method of reconstruction based on stereo matching is adopted in this paper,including camera calibration,stereo correction,target detection and matching,point cloud reconstruction and so on.For two different objects,the Hough gradient algorithm based on optimized sorting and the least squares ellipse fitting algorithm based on contour extraction are used to find the center of the feature points.Then the affine transform matching algorithm based on epipolar constraint is used for stereo matching.Finally,the 3D modeling and surface reconstruction are performed to obtain the structural morphology and surface information of the spatial object.In temporal series reconstruction,the method of reconstruction based on tracking is adopted in this paper,including target tracking,2D trajectory fitting and repair,3D trajectory reconstruction and so on.For the image sequence,the linear dynamic model algorithm based on pyramid optical flow is used to track the object.The spatial coordinates and motion parameters of the feature point are achieved by the motion trajectory reconstruction.The experimental results show that in spatial reconstruction,the proposed matching method is more accurate than the traditional stereo algorithm.Meanwhile,in temporal series reconstruction,the error of the target tracking is reduced.Finally,the 3D motion trajectory of the object and the motion pattern at a certain moment shows that the proposed method in this paper can obtain ideal results,which proving the effectiveness of this method.