基于图像序列的波浪三维重建算法优化  

Optimization of wave 3D reconstruction algorithm based on image sequence

作　　者：郑耿忠 宋亚男[1] 徐荣华[1] 黎兆锴 李发义 Zheng Gengzhong;Song Yanan;Xu Ronghua;Li Zhaokai;Li Fayi(School of Automation Guangdong,Guangdong University of Technology,Guangzhou 510006,China)

机构地区：[1]广东工业大学自动化学院,广东广州510006

出　　处：《无线互联科技》2022年第18期135-138,共4页Wireless Internet Technology

基　　金：广东省自然科学基金,项目名称:船舶动力定位系统安全性动态建模与评估研究,项目编号:2018A030313775。

摘　　要：文章使用Pixel2Mesh对波浪进行三维重建,其中针对重建结果精度低的问题,利用波浪图像序列的时间连续性获得的先验知识以提高重建表面的精度;采用插值结合卷积神经网络的方式,加强重建表面的细节。实验表明,优化后算法能重建出更高精度的波浪表面。In this paper,3 D reconstruction of waves is based on Pixel2 Mesh,aiming at the problem of low accuracy of reconstruction results,the prior knowledge obtained from the time continuity of wave image sequence is used to improve the accuracy of reconstructed surface.Interpolation and convolutional neural network were used to reconstruct surface details.Experiments show that the optimized algorithm can reconstruct the wave surface with higher precision.

关 键 词：波浪 三维重建 插值算法 卷积神经网络 图像序列 

分 类 号：P731.22[天文地球—海洋科学] TP391.41[自动化与计算机技术—计算机应用技术]