用于面部虚拟整形的三维人脸重建系统设计

为解决面部整形手术成功率低的问题,设计了用于面部虚拟整形的三维面部重建系统。该系统首先利用高精度深度相机采集人脸多角度深度图像,将每幅深度图像分别转换为对应点云模型并进行去噪预处理。接着,将正面点云和侧面点云进行粗匹配实现点云拼接,并提出一种改进的精配准算法进行精配准,得到目标人脸三维点云模型。对点云模型网格化处理后,提出一种基于半边结构的细分算法来提高重建模形精度,最后完成贴图即可构建人脸模型。为了验证所设计系统的重建精度,设计了对比实验。结果表明:系统重构人脸与其他重建系统相比,提高了人脸精度,且与真实人脸间差距小于2 mm,可以将其应用于面部虚拟整形,从而对专业医师提供术前指导,提高手术成功率。

3D Face Reconstruction System for Face Virtual Shaping

In order to solve the problem of low success rate of facial plastic surgery,a 3D face reconstruction system for facial virtual plastic surgery was designed. Firstly,the system used a high-precision depth camera to collect multi-angle depth images of human face,each depth image was converted to corresponding point cloud model. The point cloud data was denoised. Then the front point cloud and two side point clouds were roughly matched to realize point cloud stitching,the system used an improved precise registration algorithm for registration and obtained the 3D point cloud model of the target face. The point cloud model was meshed,a subdivision algorithm with half-edge structure was proposed to improve the accuracy of remodeling,finally the face model could be constructed by completing the texture. In order to verify the reconstruction accuracy of the designed system, a comparative experiment is designed. The experimental result shows that the system compared with other reconstruction systems,the system"s face reconstruction improves the accuracy of the face, and the difference between the face and the real face is less than 2mm,the application of this method in facial virtual plastic surgery can provide preoperative guidance for professional doctors and improve the success rate of surgery.