我无法对常规点云进行三角测量

时间:2016-10-20 09:15:42

标签: python c++ opencv qhull

我很容易细分两个参数化3D表面(两个凸面)。

这是两个镶嵌细分的参数曲面:

enter image description here

现在,我的目的是将两者结合在一起。但是我得到了这个:

enter image description here enter image description here

我使用Qhull来创建Delaunay三角剖分,它似乎适用于第一个凸面,但不适用于背面。 :(

这是我目前的代码(部分取自ZivS)

#include "Qhull.h"

using namespace orgQhull;

void myQhull::Qhull::runQhull3D(const pcl::PCLPointCloud2& pointCloud, const char* args)
{
    std::cout << "runQhull vertices" << std::endl;
    numVertices = 0;
    std::stringstream verticesSS;
    m_externalPoints = new PointCoordinates(3,"");  //3 = dimension
    pcl::PointCloud<pcl::PointXYZ>::Ptr cloud(new pcl::PointCloud<pcl::PointXYZ>);
    pcl::fromPCLPointCloud2(pointCloud, *cloud);
    std::vector<double> allPoints;

    for (unsigned int i = 0; i < cloud->size(); i++) {
        allPoints.push_back(cloud->at(i).x);
        allPoints.push_back(cloud->at(i).y);
        allPoints.push_back(cloud->at(i).z);
        verticesSS << cloud->at(i).x << " " << cloud->at(i).y << " " << cloud->at(i).z << "\n";
        numVertices++;
    }
    vertices += verticesSS.str();


    m_externalPoints->append(allPoints); //convert to vector<double>
    runQhull(*m_externalPoints, args);
}


void myQhull::Qhull::runQhull(const PointCoordinates &points, const char *qhullCommand2)
{
    std::string triangles;
    std::stringstream ss;
    numSimplices = 0;
    int numFaces = 0;

    std::cout << numVertices << std::endl;
    std::cout << "runQhull facets" << std::endl;

    orgQhull::Qhull qHull(points.comment().c_str(), points.dimension(), points.count(), &*points.coordinates(), qhullCommand2);
    QhullFacetList facets = qHull.facetList();
    for (QhullFacetList::iterator it = facets.begin(); it != facets.end(); ++it)
    {
        if (!(*it).isGood()) continue;
        QhullFacet f = *it;
        QhullVertexSet vSet = f.vertices();
        auto coord = f.hyperplane().coordinates();

        numFaces = vSet.size();

        ss << numFaces;
        for (QhullVertexSet::iterator vIt = vSet.begin(); vIt != vSet.end(); ++vIt)
        {
            QhullVertex v = *vIt;
            QhullPoint p = v.point();
            double * coords = p.coordinates();
            ss <<  " " << p.id() << " ";
        }
        ss << "\n";
        numSimplices++;


    }

    simplices +=  ss.str();
    std::cout << numSimplices << std::endl;

}

void myQhull::Qhull::saveOff(std::string file)
{
    std::cout << "Saving qhull.off" << std::endl;
    std::ofstream offFile;
    offFile.open(file);

    offFile << "OFF\n";
    offFile << numVertices << " " << numSimplices << " 0";
    offFile << vertices;
    offFile << simplices;
    offFile.close();

}


void myQhull::Qhull::run(const pcl::PCLPointCloud2& pointCloud)
{
    Qhull qhull;
    qhull.runQhull3D(pointCloud, "Qt");
    qhull.saveOff("qhull.off");

}

另外,我使用了OpenCV的greedy_projection但没有任何成功。它只能在不加入它们的情况下执行两个曲面细分。

知道为什么会这样吗?

1 个答案:

答案 0 :(得分:0)

最后我找到了解决方案。 将“d”添加到qhull.runQhull3D(pointCloud, "d Qt")会为上下表面生成正确的Delaunay三角剖分。 因此,因为它们是常规网格I,所以手动创建从两个表面连接顶点的边。 谢谢。