Three-dimensional human body reconstruction method and system based on local point cloud geometric information

• Main Authors: PENG WEILONG, XIAO YATIE, HUANG BAOREN, CHEN SIYUAN, TANG KEKE, TAN HENGLIANG, GAO XIN, WANG YANGTAO
• Format: Patent
• Language: Chinese; English
• Published: 09.07.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS

The embodiment of the invention provides a three-dimensional human body reconstruction method and system based on local point cloud geometric information, and the method comprises the steps: obtaining human body three-dimensional point cloud data and an SMPL skeleton matched with the human body three-dimensional point cloud data frame by frame, and carrying out the calculation according to the point cloud data to obtain local point cloud geometric information, obtaining a point cloud coordinate and a posture condition vector of each part of the human body based on the SMPL skeleton, and calculating to obtain a standard point cloud coordinate according to the SMPL skeleton and the point cloud coordinate; fusing the geometric information of the local point cloud, the posture condition vector and the standard point cloud coordinates as input data, inputting the input data into a symbolic distance neural network, training the symbolic distance neural network through various loss functions, obtaining all points wi