改进的多尺度点云自编码器网络  

Improved Fine-Grained Point Cloud Auto-encoder Network

作　　者：朱映韬 陈建[1,2] 万杰[1] 黄炜[1] 杜天放 ZHU Yingtao;CHEN Jian;WAN Jie;HUANG Wei;DU Tianfang(College of Advanced Manufacturing,Fuzhou University,Quanzhou 362200,China;College of Physics and Information Engineering,Fuzhou University,Fuzhou 350108,China)

机构地区：[1]福州大学先进制造学院,福建泉州362200 [2]福州大学物理与信息工程学院,福建福州350108

出　　处：《电视技术》2023年第11期22-25,48,共5页Video Engineering

摘　　要：三维数据的应用场景持续发展,对点云数据高效传输的需求不断增加。然而,如何压缩稀疏、非结构化、高精度的三维点以实现高效通信,仍是一个具有挑战性的问题。为此,通过改进多尺度自编码网络框架以优化局部特征提取,控制模型复杂度和计算开销;通过引入多尺度特征提取块,将潜在表征自适应地分解为多个不同尺度的局部特征;通过残差级联结构减小全局特征损失,实现特征的多尺度融合。The application scenarios of 3D data continue to evolve,and the demand for efficient transmission of point cloud data continues to increase.However,how to compress sparse,unstructured,high-precision three-dimensional points for efficient communication remains a challenging problem.In this paper,the multi-scale self-coding network framework is improved to optimize local feature extraction and control model complexity and computational overhead.By introducing multi-scale feature extraction blocks,the potential representation is decomposed into multiple local features of different scales.The residuals cascade structure is used to reduce the global feature loss and realize multi-scale feature fusion.

关 键 词：点云压缩 细粒度 稀疏卷积 端到端学习 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]