基于三维点云的重建算法

Reconstruction Algorithm Based on 3D Point Cloud

近年来,通过三维扫描仪获得点云数据越来越流行,基于点云的重建技术成为计算机图形学等领域的研究热点。首先通过KNN算法计算出每个点需要扩展为圆盘的半径大小,然后将点云数据中将每个点扩展为具有一定半径大小的圆盘来弥补点与点之间的空洞;由于圆盘半径过大会导致圆盘之间相互交叉,在圆盘上施加高斯滤波核函数来消除面元之间相互交叉带来的失真现象,同时在指定参数阈值范围内自适应地将核函数的能量限制在面元范围内,得到良好的绘制效果。

In recent years, It is more and more popular to get point cloud data from 3D scanner, As a result, point cloud based reconstruction has widely been studied in areas such as computer graphics and etc. This article uses KNN algorithm to calculate the radius of each point which needs to be extended to the disk size, then extends each point as a surfel with a certain size radius to cover the holes between each point, but the resulting problems is that the excessive length radius would produce the overlap with each surfel. We apply a gauss filter kernel function on each surfel to eliminate the cross distortion. At the same time, within the user defined parameters, we constraint the kernel function energy onto the surfel and obtain the desirable rendering result.