基于服装原型的三维服装重建方法

3D Garment Reconstruction Method Based on Garment Prototype

提出了一种基于服装原型的无接触三维服装重建方法。通过半球式拍摄获取不同空间系列视图的服装图像,经过提取特征、特征匹配、三维注册和几何验证等过程,获得初始数据集。采用多视角立体视觉技术,重构了稠密的点云模型。同时,为获得完整的3D服装点云模型,提出了一种分离3D服装点云与背景的方法。该重建技术能够高效地从一系列图像中恢复目标的点云数据,成功再现了服装在空间中的复杂动态结构,显著增强了三维服装重建模型的逼真度。实验表明:该方法可得到较完整和高质量的3D服装模型,验证了算法的有效性和可行性。

The contactless 3D clothing reconstruction method based on clothing prototype is proposed.The clothing images of different spatial series views are obtained by hemispherical shooting,and the initial data set is obtained through the process of fea-ture extraction,feature matching,3D registration and geometric verification.The dense point cloud model is reconstructed using multi-view stereo vision technology.At the same time,in order to obtain a complete 3D clothing point cloud model,a method to sep-arate 3D clothing point cloud and background is proposed.The reconstruction technique could recover the point cloud data of the tar-get from a series of images efficiently,successfully reproducing the complex dynamic structure of the garment in space and signifi-cantly enhancing the fidelity of the 3D garment reconstruction model.Experiments show that the method can obtain a relatively com-plete and high-quality 3D clothing model,which verifies the effectiveness and feasibility of the algorithm.