Journal of Frontiers of Computer Science and Technology ›› 2020, Vol. 14 ›› Issue (4): 657-668.DOI: 10.3778/j.issn.1673-9418.1903033

Previous Articles     Next Articles

Recoverable 3D Point Cloud Compression Algorithm Based on Vector Similarity

ZHANG Xukang, NIU Baoning, ZHANG Jinwen   

  1. College of Information and Computer, Taiyuan University of Technology, Taiyuan 030000, China
  • Online:2020-04-01 Published:2020-04-10

向量相似度可复原三维点云压缩算法

张旭康牛保宁张锦文   

  1. 太原理工大学 信息与计算机学院,太原 030000

Abstract:

Data compression for 3D point cloud suffers from several problems, including difficult to preserve the feature details, over-compress, and difficult to restore a compressed model. To alleviate these problems, CVS (compression based on vector similarity), consisting of a compression algorithm and a restoration algorithm based on vector similarity is proposed. This paper proposes L3A (one length, three angles) for vector similarity metrics. CVS takes a 3D point as a 3D vector connecting its coordinates and origin. According to the reading order of the 3D coordinate points, reference vectors are selected to form sampling areas covering the entire point cloud, and each area is compressed independently. In the sampling area, the least squares surface fitting algorithm is used to perform surface fitting on the point cloud including the curve, the curvature threshold is set to eliminate the coordinate points, and the surface equation parameters are stored for restoration. By controlling the length and angle changing thresholds in the L3A vector similarity, the compression rate of the dense point cloud region is higher than that of the non-dense region, and by controlling the curvature threshold, the compression ratio of the low curvature region is higher than that of the high curvature region. The compression ratio preserves the model’s detail features to the utmost extent. CVS uses the recovery information generated by the compression phase to generate a point cloud to restore the detailed features of the model, making the model features more obvious.

Key words: point cloud compression, point cloud restoration, vector similarity

摘要:

针对三维点云数据压缩中细节特征不易保留,模型平缓部位存在过度压缩以及压缩后的点云模型不易复原等问题,提出一种基于向量相似度的三维点云压缩算法和复原算法CVS。向量相似性度量采用提出的L3A进行度量。CVS把每个三维坐标点看作是连接其坐标和原点的三维向量,按照三维坐标点的读入顺序选取参考向量,生成覆盖整个点云区域的采样区域,进行分区压缩。在采样区域中使用最小二乘曲面拟合算法对包含其中的点云进行曲面拟合,设置曲率阈值剔除坐标点,并存储曲面方程参数用于复原。通过控制L3A向量相似度中的长度和角度的变化阈值,使得密集点云区域的压缩率高于非密集区域的压缩率,通过控制曲率阈值,使得低曲率区域的压缩率高于高曲率区域的压缩率,最大程度保留模型细节特征。CVS使用压缩阶段产生的复原信息生成点云来恢复模型的细节特征,使得模型特征更加明显。

关键词: 点云压缩, 点云复原, 向量相似度