[1]魏雪霞,徐增波*,王巧丽.基于二维图像的参数化人体建模[J].服装学报,2023,8(01):24-30.
　WEI Xuexia,XU Zengbo*,WANG Qiaoli.Parametric Human Modeling Based on Two-Dimensional Image[J].Journal of Clothing Research,2023,8(01):24-30.
点击复制

基于二维图像的参数化人体建模()
分享到：

《服装学报》[ISSN:2096-1928/CN:32-1864/TS]

卷:
第8卷
期数:
2023年01期
页码:
24-30
栏目:
服装智造
出版日期:
2023-02-28

文章信息/Info

Title:
Parametric Human Modeling Based on Two-Dimensional Image
作者:
魏雪霞;  徐增波*;  王巧丽
上海工程技术大学 纺织服装学院,上海 201620
Author(s):
WEI Xuexia;  XU Zengbo*;  WANG Qiaoli
School of Textiles and Fashion, Shanghai University of Engineering Science, Shanghai 201620, China
分类号:
TS 941.2
文献标志码:
A
摘要:
为解决现有基于二维图像的三维人体建模容易出现误差、模型参数估计不准确等问题,结合深度学习和优化后的人体建模算法,并利用深度学习算法提供的结果作为先验,缓解优化算法对参数初始值敏感的问题,以实现针对单帧RGB图像和多帧RGB图像的三维人体建模。同时基于HMR深度学习模型预测SMPL-X模型的初始参数,通过添加人体轮廓、二维关键点等对模型参数进一步优化求解,并利用视频序列的帧间连贯性对视频帧中人物的三维姿态进行约束。结果表明:添加人体分割轮廓约束可使重建模型更加贴合人体,能够提升模型拟合的精度; 基于视频序列帧间连贯性重建的人体模型,可减少身体旋转、四肢弯曲等误差出现的可能性,与目标对象姿态更加相近。
Abstract:
The existing 3D human modeling based on two-dimensional images is prone to errors and inaccurate estimation of model parameters. In order to alleviate the problem that the optimization algorithm is sensitive to the initial value of parameters, the results provided by the deep learning algorithm are used as a priori, and the human modeling algorithm based on deep learning and optimization is combined. This enables 3D human modeling for single frame RGB images and multi frame RGB images. The initial parameters of SMPL-X model are predicted based on HMR deep learning model, and the model parameters are further optimized by constraints such as human contour and two-dimensional key points. The inter frame coherence of the video sequence is used to constrain the 3D pose of the characters in the video frame. The results showed that adding human segmentation contour constraints could make the reconstructed model more fit the human body and improve the accuracy of model fitting. The pose of the human model reconstructed based on the inter frame coherence of the video sequence was more similar to that of the target object, which reduced the possibility of errors such as body rotation and limb bending.

参考文献/References:

[1] 顾佳雯, 周晓琪, 王军. 三维人体建模方法研究与应用[J]. 纺织科技进展, 2021(9): 15-18.
GU Jiawen, ZHOU Xiaoqi, WANG Jun. Research and application of 3D human body modeling[J]. Progress in Textile Science and Technology, 2021(9): 15-18.(in Chinese)
[2] 王巧丽, 徐增波, 田丙强. 虚拟试衣中个性化三维人体建模的研究进展[J]. 纺织导报, 2020(7): 79-82.
WANG Qiaoli, XU Zengbo, TIAN Bingqiang. Research progress of personalized 3D human body modeling in virtual fitting[J]. China Textile Leader, 2020(7): 79-82.(in Chinese)
[3] 劳秀霞, 柯略. 基于3Dmax的三维动画角色建模探析[J]. 科技传播, 2011, 3(4): 236-237.
LAO Xiuxia, KE Lüe. Analysis of 3D animation char-acter modeling based on 3Dmax[J]. Public Communication of Science and Technology, 2011, 3(4): 236-237.(in Chinese)
[4] 吕珍, 谢倩, 祝双武, 等. OpenGL导入3DS模型模拟服装试穿[J]. 西安工程大学学报, 2014, 28(4): 440- 444.
LYU Zhen, XIE Qian, ZHU Shuangwu, et al. Importing 3DS models to simulate garment fitting by OpenGL[J]. Journal of Xi’an Polytechnic University, 2014, 28(4): 440- 444.(in Chinese)
[5] 刘雁, 耿兆丰. 三维人体及服装建模方法分析与实现[J]. 计算机应用与软件, 2004, 21(10): 47- 48, 95.
LIU Yan, GENG Zhaofeng. Analysis and realization of 3D mannequin and garment modeling[J]. Computer Applications and Software, 2004, 21(10): 47- 48, 95.(in Chinese)
[6] 韩凯, 庞宗强, 王龙, 等. 基于深度扫描仪的高辨识度三维人体模型重建方法[J]. 图学学报, 2015, 36(4): 503-510.
HAN Kai, PANG Zongqiang, WANG Long, et al. High identification 3D human body model reconstruction method based on the depth scanner[J]. Journal of Graphics, 2015, 36(4): 503-510.(in Chinese)
[7] GUAN P, WEISS A, B?LAN A O, et al. Estimating human shape and pose from a single image[C]//2009 IEEE 12th International Conference on Computer Vision. Kyoto:IEEE, 2010: 1381-1388.
[8] BOGO F, KANAZAWA A, LASSNER C, et al. Keep it SMPL: automatic estimation of 3D human pose and shape from a single image[M]//Computer Vision-ECCV 2016. Berlin: Springer International Publishing, 2016: 561-578.
[9] 张广翩, 计忠平. 基于二维点云图的三维人体建模方法[J]. 计算机工程与应用, 2020, 56(19): 205-215.
ZHANG Guangpian, JI Zhongping. Method of 3D human body modeling based on 2D point cloud image[J]. Computer Engineering and Applications, 2020, 56(19): 205-215.(in Chinese)
[10] CAO Z, SIMON T, WEI S H, et al. Realtime multi-person 2D pose estimation using part affinity fields[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Honolulu: IEEE, 2017: 1302-1310.
[11] GONG K, GAO Y M, LIANG X D, et al. Graphonomy: universal human parsing via graph transfer learning[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). Long Beach: IEEE, 2020: 7442-7451.
[12] KANAZAWA A, BLACK M J, JACOBS D W, et al. End-to-end recovery of human shape and pose[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 7122-7131.
[13] PAVLAKOS G, CHOUTAS V, GHORBANI N, et al. Expressive body capture: 3D hands, face, and body from a single image[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). Long Beach: IEEE, 2020: 10967-10977.
[14] MOULON P, MONASSE P, PERROT R, et al. OpenMVG: open multiple view geometry[M]//Reproducible Research in Pattern Recognition. Berlin: Springer International Publishing, 2017: 60-74.
(责任编辑:张 雪)

更新日期/Last Update: 2023-02-28