摘要： 提出一种基于统一计算设备架构(CUDA)的双边滤波点云去噪算法，将点云去噪划分为多个并行度较高的步骤，利用GPU的并行计算能力，设计每个步骤的CUDA核函数。采用高斯加权的法矢计算方法，在双边去噪算法中加入面积权重缓解过光顺。实验结果表明，该算法能有效提高法矢计算的准确度，与CPU算法相比，计算速度提高了多个数量级。

关键词: 统一计算设备架构, GPU并行计算, 点云去噪, 双边滤波

Abstract: This paper proposes a Compute Unified Device Architecture(CUDA)-based improved bilateral filtering point cloud denoising algorithm. The point cloud denoising algorithm is divided into several steps in a very high degree of parallelism. It separately designs CUDA kernel functions for each step, effectively emploies the GPU’s parallel computing power. It uses Gaussian-weighted method of calculating normal vector and effectively improves the accuracy of normal vector calculation. The bilateral denoising algorithm adds the weight of the area to alleviate the excessive smoothing. Experimental results show that the algorithm is stable and efficient, faster than the CPU calculation of multiple orders of magnitude.

Key words: Compute Unified Device Architecture(CUDA), GPU parallel computing, point cloud denoising, bilateral filtering

中图分类号: 

• TN911.73