| 429 | 6 | 126 |
| 下载次数 | 被引频次 | 阅读次数 |
针对深度相机连续移动中，由于位姿误差积累导致所感测的人体3D点云模型精度下降的问题，研究提出一种Surfel-BA三维重建优化算法，通过引入全局BA优化，提高了位姿估计与稠密重建的Surfel表面精度，从而实现基于感测的人体点云精确模型。基于TUM数据集的位姿精度比测结果表明，新算法提升了ATE误差统计特性、有效抑制RPE误差，进而设计实现基于感测的数字化人台与3D虚拟旗袍模型，并搭建Kinect与Unity3D虚拟试穿集成环境，开展真人旗袍虚拟试穿现场实验，获得交互性好、贴体逼真的体验效果，且快速便捷、成本低，广泛适用于各种数字时尚应用场景。
Abstract：In order to avoid decreasing accuracy of the 3 D point-cloud human model due to the pose error accumulation during a continuous movement of the depth-camera, a novel Surfel-BA algorithm for 3 D reconstruction is proposed by introducing a global BA optimization method. The accuracy of both pose estimation and dense fusion of the surface element can be improved, so as to realize the accurate fitting human point-cloud model through camera sensing. The comparison results testing on TUM data set show that the new algorithm can improve the statistical characteristics of ATE error and effectively suppressing RPE error. Furthermore, both digital mannequin model and 3 D virtual cheongsam model of the fitting person were constructed, and the cheongsam virtual fitting experiment was carried out with Kinect and Unity3 D. It turns out that the proposed virtual cheongsam modeling system can give a realistic experience on the virtual fitting with advantages of good interaction, convenience and low cost. It is suitable for various digital fashion applications.
[1] 黎锋，刘雅玲.基于顾客视角的虚拟试衣系统发展及对策研究[J].天津纺织科技，2021(2):23-27.
[2] 亚马逊计划推出虚拟试衣间Outfit-VITON,跨境达人[EB/OL].[2022-03-10].https://www.hcggzy.cn/kuajing/dianshang/7406.html.
[3] LIU T,XU H,ZHANG X W.3D Clothing Transfer in Virtual Fitting Based on UV Mapping[J].2018 11th International Congress on Image and Signal Processing,Bio Medical Engineering and Informatics (CISP-BMEI),Beijing,China,2018:1-6
[4] 中国得物公司官网，https://www.dewu.com/(阅读时间2022.2.10)
[5] 薛静雅.基于Kinect的虚拟试衣系统设计[D].呼尔浩特：内蒙古大学，2018.
[6] 张卓，丛洪莲.基于虚拟现实技术的服装3D仿真与应用[J].上海纺织科技，2021,49(5):19-25.
[7] 胡佳琪，宋莹.基于CLO3D虚拟试衣技术的旗袍着装效果评价研究[J].丝绸，2021,58(12):73-79.
[8] 冀艳波，王玲丽，刘凯旋.基于数字化三维人体模型的旗袍定制设计[J].纺织学报，2021,42(1):133-137.
[9] 朱庆艳.EMTM背景下生活装旗袍的样板自动生成系统研究[D].杭州：浙江理工大学，2017.
[10] 吴建清，宋修广.同步定位与建图技术发展综述[J].山东大学学报(工学版),2021,51(5):16-31.
[11] NEWCOMBE R A,IZADI S,HILLIGES O,et al.Kinect Fusion:Real-time Dense Surface Mapping and Tracking[C]//2011 10th IEEE International Symposium on Mixed and Augmented Reality.Basel,Switzerland.IEEE,2012:127-136.
[12] WHELAN T,LEUTENEGGER S,RF Salas-Mor-eno,et al.Elastic Fusion:Dense SLAM Without a Pose Graph[C].Robotics:Science & Systems,2015.
[13] 张震，郑宏，周璇，等.一种结合ORB特征和视觉词典的RGB-D SLAM算法[J].计算机工程与应用，2018,54(12):170-176.
[14] 谷晓琳，杨敏，张燚，等.一种基于半直接视觉里程计的RGB-D SLAM算法[J].机器人，2020,42(1):39-48.
[15] XU H T,XU J M,XU W W.Survey of 3D Modeling Using Depth Cameras[J].Virtual Reality & Intelligent Hardware,2019,1(5):483-499.
[16] 黎博文，王萍，刘玉叶.基于人体动态特征的三维服装虚拟试穿技术[J].纺织学报，2021,42(9):144-149.
[17] 罗伟，梁世豪，姜鑫，等.基于微软Kinect的体感控制智能窗帘系统[J].微型电脑应用，2020,36(3):64-68.
基本信息:
中图分类号:TS941;TP391.41
引用信息:
[1]徐凤仪,王萍,黎博文,等.基于Kinect人体点云三维重建的个性化旗袍虚拟试穿[J].微型电脑应用,2022,38(10):108-112+116.
2022-10-20
2022-10-20