摘要
三维重建的研究是近年来计算机视觉领域的一个热点。基于单个深度相机的KinectFusion算法具有实时性高、比较廉价等优点,在KinectFusion算法基础上提出了能够自动识别人脸的有效区域,通过获取相应的深度数据和RGB数据来进行带纹理的人脸三维重建方法。在点云配准阶段引入一个惩罚因子来平滑人脸的非刚体运动。根据获取的颜色信息建立颜色信息的场函数并进行纹理贴图,得到更加逼真的带纹理的三维模型;在重建过程中使用了GPU加速实现实时性。
3D reconstruction is a hot spot in the field of computer vision. Based on a depth sensor characteristics of real time and low cost of Kinect Fusion algorithm,a method which can automatically recognize face region by obtaining depth data and RGB data to reconstruct the 3D model on the basis of Kinect Fusion. This method adds a penalty factor to smooth the motivation of human face. Texture integration is applied to get a color information.GPU acceleration is applied during the reconstruction.
出处
《传感器与微系统》
CSCD
2017年第8期126-129,共4页
Transducer and Microsystem Technologies
基金
国家自然科学基金资助项目(61573356)
关键词
人脸三维重建
纹理融合
GPU加速
human face reconstruction
texture integration
GPU acceleration
