Research Article
Mesh2Measure: A Novel Body Dimensions Measurement Based on 3D Human Model
@INPROCEEDINGS{10.1007/978-3-030-99188-3_6, author={Tao Song and Rui Zhang and Yukun Dong and Xixi Tao and Hongcui Lu and Baohua Liu}, title={Mesh2Measure: A Novel Body Dimensions Measurement Based on 3D Human Model}, proceedings={Intelligent Technologies for Interactive Entertainment. 13th EAI International Conference, INTETAIN 2021, Virtual Event, December 3-4, 2021, Proceedings}, proceedings_a={INTETAIN}, year={2022}, month={3}, keywords={Anthropometric dimensions 3D Human reconstruction Dense elliptic model}, doi={10.1007/978-3-030-99188-3_6} }- Tao Song
Rui Zhang
Yukun Dong
Xixi Tao
Hongcui Lu
Baohua Liu
Year: 2022
Mesh2Measure: A Novel Body Dimensions Measurement Based on 3D Human Model
INTETAIN
Springer
DOI: 10.1007/978-3-030-99188-3_6
Abstract
In this work, we propose an anthropometric dimensions measurement method based on 3D human model, namely Mesh2Measure. In our method, Human body features in the front and side images are firstly extracted and fused. And then, the feature vectors are attached to the template mesh model SMPL by using Graph-CNN, and 3D coordinates of the model vertices are regressed. Anthropometric dimensions of height, length, width and depth are calculated by scale conversion based on the model vertex coordinates. A novel general dense elliptic model is developed for the curve dimension or closed circumference dimension, which obtains human body dimensions by accumulating the length of elliptic segments with different coefficients. Data experiments are conducted by measuring 100 subjects. Experimental results show that our Mesh2Measure model can measure 38 main dimensions of human body in 15 s, and more importantly, the accuracy rate is 97.4% compared with the ground truth dimensions by manual measurements.
