1. 基于角度半径的青年男子躯干部形态相似性匹配.
- Author
-
盛锡彬 and 顾冰菲
- Abstract
With the continuous development of the world economy and people's living level, the demand for clothing comfort and fitness has become increasingly prominent. In the process of online clothing consumption, it is difficult for consumers to try on and choose suitable clothes. How to analyze the human body shape quickly and accurately is an urgent problem to be solved. Increasing attention has been paid to non-contact two-dimensional anthropometry, with the recognition and matching of human body contour being the key research direction. Many domestic and foreign scholars have carried out substantial studies, and put forward a variety of methods such as RBF neural network, cubic spline function, least square method and Fourier coefficient, but these methods can not fuzzy match the human body shape. The shape context method has also been used for shape matching. Although this method can improve the accuracy and speed of matching, it is still difficult to match human form with the shape context. Therefore, this study focused on how to describe human body contour shape quantitatively based on three-dimensional point cloud data of human body. An angular radium-based contour description method was proposed. One hundred and eighty-seven young male college students aged 18-25 were selected as research objects, and three-dimensional point cloud data were obtained by three-dimensional scanning method. Then the reverse engineering software was used to identify the feature points of the human trunk and intercept the silhouette of the front and lateral trunk. The frontal and lateral trunk coordinate systems were established respectively, and 27 frontal and 60 lateral angular radii were extracted to establish the frontal and lateral trunk form angular radii database based on three-dimensional human point cloud. In order to obtain the human body with a similar shape, the angular radius was used as the reference to judge whether the cross-section shape of the characteristic parts was consistent, and the similarity was quantified by the root mean square of the shape description index. Finally, the angular radius value extracted by 14 subjects was taken as the object to be matched, and the root-mean- square error method was used to achieve the similarity matching of the torso shape of the three-dimensional point cloud cross-section. The effectiveness of the method was verified by the MatchSharp operator in the OpenCV framework. The results of the two matching methods showed that the 13 subjects matched the frontal and lateral sections of the trunk with an accuracy of 92.86%. To further prove the accuracy of the matching method, the measurement parameters of human characteristic parts were tested by T test. The results showed that the error range of the ratio of width and thickness was between 0.053 and 0.48, and the error of the angle ranged from 3.29° to 2.73°, with Sig. values being greater than 0.05. Therefore, the matching method proposed in this study has certain feasibility. This study provides a new way for human body shape similarity matching, theoretical basis and technical support for human body 3D point cloud based morphological analysis and 3D reconstruction research, and can be further used for 3D virtual reconstruction based on human body photos. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF