摘要
基于图像的三维建模方法的代表PMVS2算法,对于纹理信息丰富的图像具有良好的效果,然而在现实中并非所有的场景图像都具有丰富纹理,如白色的墙壁等,这些场景图像中相似的特征区域较多,导致特征匹配点对中存在大量的噪声信息,从而影响到三维建模的精度。提出的方法对产生三维重构噪声的原因进行分析,提出一种简单实用的纹理添加方式,以及相机矩阵优化和建模扩展滤波优化方法,有效地减少了弱纹理图像三维建模中的噪声点云,提高了建模完整性和精度。
The representative image based 3D modeling PMVS2 algorithm has a good effect on the rich texture information images. However, not all the shoot object images are always with rich textures, such as white wall. Because these scenes have many similar characteristics which will lead to much noise information in feature match, the accuracy of 3D modeling will be finally affected. The reasons of weak texture 3D reconstruction noise were analyzed. Then, a simple and practical method which added textures for these objects, and a method for camera matrix optimization, modeling expansion and filter were put forward, which have reduced the noise of weak texture image effectively, and have improved the completeness and accuracy of modeling.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2013年第10期2449-2453,2459,共6页
Journal of System Simulation
关键词
弱纹理
图像三维建模
相机矩阵优化
扩展滤波
weak texture
image based modeling
camera matrix optimization
expend and filter
