基于特征信息分类的三维点数据去噪

为了有效去除获取三维点云数据时的噪声,同时又不损失模型的特征信息,提出了一种基于三维点云特征信息分类的去噪算法。首先采用主成分分析法和二次曲面拟合法估算三维点云的微分几何信息;然后根据点云平均曲率的局部特征权值,将点云数据划分为特征信息较少的平坦区域和特征信息丰富的区域,针对不同特征区域分别采用邻域距离平均滤波算法和自适应双边滤波算法进行去噪滤波。实验结果表明:滤波后点云数据的最大误差为0.144 7mm,标准偏差为0.021 0mm。在不同噪声强度下,该去噪算法均能够达到较好的去噪效果,并保留点云的高频特征信息。

Denoising of three-dimensional point data based on classification of feature information

To ensure no loss of feature information of model and effectively eliminate the noise at the time of acquisition of three-dimensional point cloud data, a kind of denoising algorithm based on classification of three-dimensional point cloud feature information was proposed. Firstly, the principal component analysis and conicoid fitting method were adopted to estimate the differential geometry information of three-dimensional point cloud. Then, according to the local feature weight of average curvature of point cloud, the cloud data was divided into flat region with little feature information and region with rich feature information. Pursuant to different feature regions, the average filtering algorithm with neighborhood distance and self-adaptive bilateral filtering algorithm were respectively adopted to perform denoising and filtering. The experimental results indicate that the maximum error of point cloud data is 0.144 7mm, and standard deviation is 0.021 0 mm after filtering. Under different noise intensities, this denoising algorithm may reach preferable denoising effects and reserve the high-frequency feature information of point cloud.