由稀疏至稠密的多图像3D重建的实现  

Implementation of 3D Reconstruction from Sparse to Dense Using Multiple Images

作　　者：钟向阳[1] 

机构地区：[1]嘉应学院计算机学院,广东梅州514015

出　　处：《嘉应学院学报》2016年第8期28-33,共6页Journal of Jiaying University

摘　　要：给出了一种由稀疏至稠密实现3D模型的重建方法.首先对采集的多幅图像进行2D特征点提取,通过基于绝对二次曲面的分层重建方法建立相关投影矩阵;然后利用标定的投影矩阵对2D特征点进行逆投影生成稀疏的3D空间点,进一步扩展3D点的邻域生成稠密点集;最后经滤波后生成精确3D目标模型.实验结果表明重建的物体可视性较高.This paper presents a method for the reconstruction of 3D model,which is realized by from sparse to dense. Firstly the 2D feature points are extracted from the collected multiple images and the relative projection matrixs are established by the method of the stratified reconstruction based on the absolute quadric. Secondly the 2D feature points are projected to generate sparse 3D space points by using the inverse projection matrixs,and the neighborhood of 3D points are extended to generate a dense set of points. Finally the accurate 3D target model is generated by filtering. Experimental results show that the reconstructed objects are highly visible.

关 键 词：三维重建 射影重建 绝对二次曲面 相机标定 面片 

分 类 号：TH181[机械工程—机械制造及自动化]