使用PCL（點雲庫）獲取對象信息的數組

Question

作為一個例子，在2D空間中，正方形的位置由'1'的集合表示，空的空間的位置由'0'的集合表示。

00000000000000
00000111000000
00000111000000
00000111000000
00000000000000

發現pcl :: OrganizedFastMesh類是一種執行此任務的方法。 http://docs.pointclouds.org/trunk/classpcl_1_1_organized_fast_mesh.html的網站顯示了該課程的詳細信息，但是，我很難理解。 例如，如果輸入文件名是'example.stl'，那么如何獲取上述'0'和'1的信息到數組'siteInfo [] []'？

Answer 1

要從stl導入，您需要從頂點生成點。 這是我將stl導入pcl點雲的功能：

//generates an evenly distributed point cloud from stl file (assumed to be scaled in mm)
//maxPPDist: desired max distance between points (all surfaces will be upsampled until this density is reached)
//normalNeighborCount: how many points to include in NN normal resampling. (should match what is used on camera cloud)
bool PCL_Util::importCAD_STL(pcl::PointCloud<pcl::PointNormal>::Ptr &objectCloud,
std::string fileName,
double maxPPDist,
bool normResample,
int normalNeighborCount,
bool fastNormRecombination)
{

pcl::PolygonMesh mesh;
int fileReadVal;
try
{
    fileReadVal = pcl::io::loadPolygonFileSTL(fileName, mesh);
}
catch (...)
{
    return false;
}

if (fileReadVal == 0)
{
    PCL_ERROR("Failed to load STL file\n");
    return false;
}
else
{
    pcl::PointCloud<pcl::PointNormal>::Ptr outputCloud(new pcl::PointCloud<pcl::PointNormal>);

    pcl::PointCloud<pcl::PointXYZ> objCloud;
    pcl::PCLPointCloud2 ptCloud2 = mesh.cloud;
    pcl::fromPCLPointCloud2(ptCloud2, objCloud);

    for (int i = 0; i < mesh.polygons.size(); i++)
    {
        pcl::Vertices currentPoly = mesh.polygons[i];

        for (int ii = 0; ii < currentPoly.vertices.size(); ii++)
        {
            pcl::PointNormal currentPt = pcl::PointNormal();
            currentPt.x = objCloud[currentPoly.vertices[ii]].x;
            currentPt.y = objCloud[currentPoly.vertices[ii]].y;
            currentPt.z = objCloud[currentPoly.vertices[ii]].z;
            outputCloud->points.push_back(currentPt);//push in points without normals
        }

        //make the assumption that at least 3 verticies for last poly (standard stl... not sure how dirty this is)
        int index = outputCloud->points.size() - 1;
        pcl::PointXYZ pt3(outputCloud->points[index].x, outputCloud->points[index].y, outputCloud->points[index].z);
        pcl::PointXYZ pt2(outputCloud->points[index - 1].x, outputCloud->points[index - 1].y, outputCloud->points[index - 1].z);
        pcl::PointXYZ pt1(outputCloud->points[index - 2].x, outputCloud->points[index - 2].y, outputCloud->points[index - 2].z);

        Eigen::Vector3f vec12(pt2.x - pt1.x, pt2.y - pt1.y, pt2.z - pt1.z);
        Eigen::Vector3f vec23(pt3.x - pt2.x, pt3.y - pt2.y, pt3.z - pt2.z);
        Eigen::Vector3f vecNorm = vec12.cross(vec23);
        vecNorm.normalize();

        for (int ii = 0; ii < 3; ii++)
        {
            outputCloud->points[index - ii].normal_x = vecNorm[0];
            outputCloud->points[index - ii].normal_y = vecNorm[1];
            outputCloud->points[index - ii].normal_z = vecNorm[2];
        }

        //interpolate each triangular surface to fit desired resolution
        if (maxPPDist != -1)
        {
            interpolateTriangle(outputCloud, maxPPDist);
        }
    }

    if (fastNormRecombination)//faster by an order of magnitude, but less accurate normals stl surface join points
    {
        voxelPruneCloud<pcl::PointNormal>(outputCloud, maxPPDist / 2.0f, maxPPDist / 2.0f, maxPPDist / 2.0f);
    }
    else//very slow, but generates more accurate normals at joint points
    {
        combineColocatedPoints(outputCloud, maxPPDist / 2.0f);
    }

    if (normResample)//uses the current normals as hemisphere guides for newly calculated normals
    {
        resampleNormalCloud(outputCloud, normalNeighborCount);
    }


    copyPointCloud(*outputCloud, *objectCloud);
    printf("File imported successfully?!\n");
    return true;
}
}

至於生成2d陣列​​...我會研究光線投射來生成結構化點雲/表面。 （ http://www.pcl-users.org/From-3D-point-cloud-to-depth-map-td4027567.html ）我親自編寫了自己的實現以允許線程化，但我確信pcl已經構建了一些在功能。