Point cloud fusion method for human body three-dimensional reconstruction

Abstract

The invention discloses a point cloud fusion method for three-dimensional reconstruction of a human body, and the method comprises the steps: carrying out the three-dimensional data collection of thehuman body standing on a turntable through employing a fixed Kinect device, obtaining the point cloud data of a plurality of angles of the human body, carrying out the rotation of the point cloud of each angle through employing the center of gravity of the point cloud as a rotation center, and obtaining the point cloud data of a unified coordinate system; calculating the normal vector and the umbrella curvature of each point in each point cloud to obtain a point cloud feature; and finally, carrying out matching fusion on the point cloud data at multiple angles according to the point data information of the same point cloud characteristics in the point clouds at different angles by adopting an ICP algorithm based on a bidirectional KD tree. According to the invention, an umbrella-shaped algorithm is adopted to calculate the point cloud curvature; the point cloud is registered by adopting a bidirectional KD tree method, so that the accuracy of feature point extraction is improved, accurate data is provided for initial iteration data of point cloud fusion, and accurate feature matching data is provided for point cloud fusion, so that a better human body three-dimensional reconstruction effect can be obtained, and the sense of reality of human body modeling is improved.

Description

technical field [0001] The invention relates to a point cloud fusion method used for three-dimensional reconstruction of the surface of a human body, and belongs to the technical field of data processing. Background technique [0002] With the continuous development of 3D reconstruction technology, in computer vision, virtual reality and other fields, obtaining object contour information and 3D modeling, 3D reconstruction technology has important applications, and the acquisition of human body contour information has more important significance. A point cloud is a collection of point data on the surface of a person or object. The accuracy of point cloud feature extraction obtained from different angles plays a decisive role in the accuracy of 3D reconstruction. Only high-precision feature extraction can obtain a closer to the real 3D reconstruction effect. [0003] The existing 3D point cloud feature extraction technology has low efficiency and poor real-time performance, a...

AI Technical Summary

Problems solved by technology

[0003] The existing 3D point cloud feature extraction technology has low efficiency and poor real-time performance, and the acquisition of feature points is not accurate enough, which affects the accuracy of subsequent point cloud registration. Point cloud registration and matching feature points are not accurate, which affects the final 3D reconstruction effect

[0004] Although the 3D human body reconstruction method based on the Kinect device can easily obtain the surface data of the human body, the obtained results are not complete. The equipment collected during the reconstruction process needs to be moved, which introduces errors for the subsequent reconstruction process. There is a large error, which reduces the realism of human body modeling

[0005] In addition, when registering the human body surface data obtained from different angles, the traditional point cloud registration method, the feature extraction accuracy is not accurate enough, and the error will be further increased when matching the feature points obtained by different angle point clouds

Images