Surface Reconstruction Parallel Study In Heterogeneous Environments

Three dimensional reconstructions in the field of computer vision is widely used in virtual reality, digital city, art, cultural relic reconstruction etc.. Surface reconstruction is an important component in the process of 3d reconstruction. This paper is based on the Poisson reconstruction method. This algorithm is based on point cloud data as input, the point cloud data set is a set of independent points in 3D space, with location information, through the calculation of generating surface model. However, due to the algorithm itself leads to low efficiency of the algorithm processing mechanism. In summary, this paper gives the Poisson reconstruction speed optimization method using OpenMP(Open Multi-Processing) and CUDA(Compute Unified Device Architecture).In this paper, the basic introduction of surface reconstruction based on research status at home and abroad, focuses on the octree structure and data mapping of Poisson surface reconstruction, and introduces the multi-core CPU(Central Processing Unit) OpenMP acceleration technology and CUDA architecture based GPU(Graphics Processing Unit) acceleration technology. Improve the processing mechanism of the original algorithm, in data processing is the independence of each layer with a certain parallelism. Model-based design of parallel multicore CPU, GPU and CPU/GPU hybrid parallel algorithm acceleration. For different data sets on different platforms to experiment to analyze the accuracy of the reconstruction, rebuilding efficiency and reconstruction scalability etc., further validates the surface reconstruction can parallelism.