摘要
针对基于视频图像的三维重建所需设备复杂和计算量大的问题,提出基于CUDA的可视外壳并行计算仿真实验平台。用三维图形绘制管线中的视点模拟真实相机,应用相机参数将各图像中物体轮廓坐标统一转换到世界坐标系,利用CUDA并行计算可视外壳。在此平台上将基于体素的可视外壳生成算法转化为CUDA线程块并行计算,仿真实验表明加速效果明显。
Due to the problems existing in image-based three-dimensional modeling, such as requirement for complex devices and heavy calculations, a simulation platform for parallel computing visual hull based on CUDA acceleration was proposed. According to 3D graphics processing pipeline, the viewport was used to simulate real cameras, and with the aid of camera parameters, the coordinates of object silhouette in each image were transformed in world coordinate system, and visual hull was parallel computed using CUDA. On this platform, the voxel-based visual hull algorithm was converted into the process of parallel CUDA thread blocks. The simulation experiment shows the simulation platform based on CUDA can obviously improve the speed of the reconstruction.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2012年第1期72-75,共4页
Journal of System Simulation
基金
中央高校基本科研业务费专项资金(09CX04034A)
国家973项目(2009CB320805)
关键词
三维重建
可视外壳
GPU通用计算
CUDA
并行计算
three-dimensional reconstruction
visual hull
general purpose GPU
CUDA
parallel computing