一种改进的具有半径滤波和RANSAC优点的点云图像去噪方法

An improved point cloud image denoising method with advantages of radius filtering and RANSAC

为了剔除点云中离散点噪声和密集平面噪声且保留点云的特征,提出了一种改进的具有半径滤波和RANSAC优点的点云图像去噪新方法。首先利用体素下采样对原始点云数据进行精简,然后针对离散点噪声,使用半径滤波将其剔除,最后在保留原始RANSAC算法的基础上,引入高度信息参数对拟合出的平面进行区分,据此进一步剔除点云中的密集平面噪声。实验结果表明,经过本方法去噪后SNR比传统方法和新方法处理后的分别高出152.94%和79.17%;同时本方法在运行时间上也分别快了96.27%和98.56%。

In order to eliminate discrete point noise and dense plane noise in point cloud and retain its characteristics, this paper proposes an improved point cloud image denoising method with the advantages of radius filtering and RANSAC. Firstly, the voxel downsampling is used to streamline original point cloud data, and then the radius filtering is used to eliminate discrete point noise. Finally, on the basis of retaining original RANSAC algorithm, the height information parameter is introduced to distinguish the fitted planes. Accordingly, the dense plane noise in point cloud is further eliminated. The experimental results show that the SNRs after denoising by this method are152.94% and 79.17% higher than these of traditional method and new method, respectively. At the same time, the running time of the method is 96.27% and 98.56% faster respectively.