PCL-基于超体聚类的LCCP点云分割

一、LCCP方法

LCCP指的是Local Convexity-Constrained Patch，即局部凸约束补丁的意思。LCCP方法的基本思想是在图像中找到局部区域内的凸结构，并将这些结构用于分割图像或提取特征。这种方法可以帮助识别图像中的凸物体，并对它们进行分割。LCCP方法通常结合了空间和法线信息，以提高图像分割的准确性和稳定性。

LCCP算法大致可以分成两个部分：1.基于超体聚类的过分割。2.在超体聚类的基础上再聚类。
该方法流程图如下：

二、代码实现

#include <iostream>
#include <pcl/ModelCoefficients.h>
#include <pcl/io/pcd_io.h>
#include <pcl/point_types.h>
#include <pcl/sample_consensus/method_types.h>
#include <pcl/sample_consensus/model_types.h>
#include <pcl/segmentation/sac_segmentation.h>
#include <pcl/visualization/cloud_viewer.h>
#include <pcl/filters/extract_indices.h>
#include <boost/thread/thread.hpp>
#include <stdlib.h>
#include <cmath>
#include <limits.h>
#include <boost/format.hpp>
#include <pcl/console/parse.h>
#include <pcl/io/pcd_io.h>
#include <pcl/visualization/pcl_visualizer.h>
#include <pcl/visualization/point_cloud_color_handlers.h>
#include <pcl/filters/passthrough.h>
#include <pcl/segmentation/supervoxel_clustering.h>
#include <pcl/segmentation/lccp_segmentation.h>
#include <vtkPolyLine.h> 
#include <pcl/point_cloud.h>
#include <pcl/segmentation/supervoxel_clustering.h>
#include <pcl/visualization/pcl_visualizer.h>


using namespace std;
typedef pcl::PointXYZ PointT;
typedef pcl::LCCPSegmentation<PointT>::SupervoxelAdjacencyList SuperVoxelAdjacencyList;
//邻接线条可视化
void addSupervoxelConnectionsToViewer(pcl::PointXYZRGBA& supervoxel_center, pcl::PointCloud<pcl::PointXYZRGBA>& adjacent_supervoxel_centers,
	std::string supervoxel_name, pcl::visualization::PCLVisualizer::Ptr& viewer)
{
	vtkSmartPointer<vtkPoints> points = vtkSmartPointer<vtkPoints>::New();
	vtkSmartPointer<vtkCellArray> cells = vtkSmartPointer<vtkCellArray>::New();
	vtkSmartPointer<vtkPolyLine> polyLine = vtkSmartPointer<vtkPolyLine>::New();


	for (auto adjacent_itr = adjacent_supervoxel_centers.begin(); adjacent_itr != adjacent_supervoxel_centers.end(); ++adjacent_itr)
	{
		points->InsertNextPoint(supervoxel_center.data);
		points->InsertNextPoint(adjacent_itr->data);
	}
	vtkSmartPointer<vtkPolyData> polyData = vtkSmartPointer<vtkPolyData>::New();
	polyData->SetPoints(points);
	polyLine->GetPointIds()->SetNumberOfIds(points->GetNumberOfPoints());
	for (unsigned int i = 0; i < points->GetNumberOfPoints(); i++)
		polyLine->GetPointIds()->SetId(i, i);
	cells->InsertNextCell(polyLine);
	polyData->SetLines(cells);
	viewer->addModelFromPolyData(polyData, supervoxel_name);
}


int main(int argc, char** argv)
{
	pcl::PointCloud<pcl::PointXYZ>::Ptr cloud(new pcl::PointCloud<pcl::PointXYZ>);
	pcl::PointCloud<pcl::PointXYZ>::Ptr cloud_filtered(new pcl::PointCloud<pcl::PointXYZ>);
	pcl::PCDReader reader;
	// 读入点云PCD文件
	reader.read("E:****.pcd", *cloud);
	cout << "Point cloud data: " << cloud->points.size() << " points" << endl;
	pcl::ModelCoefficients::Ptr coefficients(new pcl::ModelCoefficients);
	pcl::PointIndices::Ptr inliers(new pcl::PointIndices);
	// 创建分割对象
	pcl::SACSegmentation<pcl::PointXYZ> seg;
	// 可选择配置，设置模型系数需要优化
	seg.setOptimizeCoefficients(true);
	// 必须配置，设置分割的模型类型、所用随机参数估计方法
	seg.setModelType(pcl::SACMODEL_PLANE);
	seg.setMethodType(pcl::SAC_RANSAC);
	seg.setDistanceThreshold(0.02);// 距离阈值 单位m。距离阈值决定了点被认为是局内点时必须满足的条件
	//seg.setDistanceThreshold(0.15);// 距离阈值 单位m。距离阈值决定了点被认为是局内点时必须满足的条件
	//距离阈值表示点到估计模型的距离最大值。
	seg.setInputCloud(cloud);//输入点云
	seg.segment(*inliers, *coefficients);//实现分割，并存储分割结果到点集合inliers及存储平面模型系数coefficients
	if (inliers->indices.size() == 0)
	{
		PCL_ERROR("Could not estimate a planar model for the given dataset.");
		return (-1);
	}
	//***********************************************************************
	//-----------输出平面模型的系数 a,b,c,d-----------
	cout << "Model coefficients: " << coefficients->values[0] << " "
		<< coefficients->values[1] << " "
		<< coefficients->values[2] << " "
		<< coefficients->values[3] << endl;
	cout << "Model inliers: " << inliers->indices.size() << endl;

	//***********************************************************************
	// 提取地面
	pcl::ExtractIndices<pcl::PointXYZ> extract;
	extract.setInputCloud(cloud);
	extract.setIndices(inliers);
	extract.filter(*cloud_filtered);

	cout << "Ground cloud after filtering: " << endl;
	cout << *cloud_filtered << std::endl;
	pcl::PCDWriter writer;
	writer.write<pcl::PointXYZ>("3dpoints_ground.pcd", *cloud_filtered, false);

	// 提取除地面外的物体
	extract.setNegative(true);
	extract.filter(*cloud_filtered);
	cout << "Object cloud after filtering: " << endl;
	cout << *cloud_filtered << endl;
	//writer.write<pcl::PointXYZ>(".pcd", *cloud_filtered, false);

	// 点云可视化
	boost::shared_ptr<pcl::visualization::PCLVisualizer>viewer0(new pcl::visualization::PCLVisualizer("显示点云"));
	//左边窗口显示输入的点云,右边的窗口显示分割后的点云
	int v1(0), v2(0);
	viewer0->createViewPort(0, 0, 0.5, 1, v1);
	viewer0->createViewPort(0.5, 0, 1, 1, v2);
	viewer0->setBackgroundColor(0, 0, 0, v1);
	viewer0->setBackgroundColor(0.3, 0.3, 0.3, v2);
	pcl::visualization::PointCloudColorHandlerCustom<pcl::PointXYZ> color_in(cloud, 255, 0, 0);
	viewer0->addPointCloud<pcl::PointXYZ>(cloud, color_in, "cloud_in", v1);
	viewer0->setPointCloudRenderingProperties(pcl::visualization::PCL_VISUALIZER_POINT_SIZE, 1, "cloud_in", v1);

	viewer0->addPointCloud<pcl::PointXYZ>(cloud_filtered, "cloud_out", v2);
	viewer0->setPointCloudRenderingProperties(pcl::visualization::PCL_VISUALIZER_COLOR, 0, 255, 0, "cloud_out", v2);
	viewer0->setPointCloudRenderingProperties(pcl::visualization::PCL_VISUALIZER_POINT_SIZE, 1, "cloud_out", v2);

	while (!viewer0->wasStopped())
	{
		viewer0->spinOnce(100);
		boost::this_thread::sleep(boost::posix_time::microseconds(1000));
	}

	//***********************************************************************
	//超体聚类 
	float voxel_resolution = 0.01f;    // 设置体素大小，该设置决定底层八叉树的叶子尺寸
	float seed_resolution = 0.15f;    // 设置种子大小，该设置决定超体素的大小
	float color_importance = 0.0f;    // 设置颜色在距离测试公式中的权重，即颜色影响超体素分割结果的比重。 真实点云都是一个颜色，所以这个参数无作用
	float spatial_importance = 0.9f;  // 设置空间距离在距离测试公式中的权重，较高的值会构建非常规则的超体素，较低的值产生的体素会按照法线
	float normal_importance = 4.0f;   // 设置法向量的权重，即表面法向量影响超体素分割结果的比重。
	bool use_single_cam_transform = false;
	bool use_supervoxel_refinement = false;

	unsigned int k_factor = 0;

	//voxel_resolution is the resolution (in meters) of voxels used、seed_resolution is the average size (in meters) of resulting supervoxels  
	pcl::SupervoxelClustering<PointT> super(voxel_resolution, seed_resolution);
	super.setUseSingleCameraTransform(use_single_cam_transform);
	super.setInputCloud(cloud_filtered); //cloud_filtered
	super.setColorImportance(color_importance);
	//Set the importance of spatial distance for supervoxels.
	super.setSpatialImportance(spatial_importance);
	//Set the importance of scalar normal product for supervoxels. 
	super.setNormalImportance(normal_importance);
	std::map<uint32_t, pcl::Supervoxel<PointT>::Ptr> supervoxel_clusters;
	super.extract(supervoxel_clusters);
	std::multimap<uint32_t, uint32_t> supervoxel_adjacency;
	super.getSupervoxelAdjacency(supervoxel_adjacency);
	pcl::PointCloud<pcl::PointNormal>::Ptr sv_centroid_normal_cloud = pcl::SupervoxelClustering<PointT>::makeSupervoxelNormalCloud(supervoxel_clusters);

	cout << "超体素分割的体素个数为：" << supervoxel_clusters.size() << endl;
	// 获取点云对应的超体素分割标签
	pcl::PointCloud<pcl::PointXYZL>::Ptr supervoxel_cloud = super.getLabeledCloud();
	pcl::visualization::PCLVisualizer::Ptr viewer1(new pcl::visualization::PCLVisualizer("VCCS"));
	viewer1->setWindowName("超体素分割");
	viewer1->addPointCloud(supervoxel_cloud, "超体素分割");
	viewer1->setPointCloudRenderingProperties(pcl::visualization::PCL_VISUALIZER_POINT_SIZE, 1, "超体素分割");
	viewer1->setPointCloudRenderingProperties(pcl::visualization::PCL_VISUALIZER_OPACITY, 0.5, "超体素分割");

	//-----------------------------------------获得体素点云的邻接单元----------------------------------------------
	multimap<uint32_t, uint32_t>SupervoxelAdjacency;
	super.getSupervoxelAdjacency(SupervoxelAdjacency);

	for (auto label_itr = SupervoxelAdjacency.cbegin(); label_itr != SupervoxelAdjacency.cend();)
	{
		uint32_t super_label = label_itr->first;//获取体素单元的标签
		pcl::Supervoxel<pcl::PointXYZ>::Ptr super_cloud = supervoxel_clusters.at(super_label);//把对应标签内的点云、体素质心、以及质心对应的法向量提取出来

		pcl::PointCloud<pcl::PointXYZRGBA> adjacent_supervoxel_centers;
		for (auto adjacent_itr = SupervoxelAdjacency.equal_range(super_label).first; adjacent_itr != SupervoxelAdjacency.equal_range(super_label).second; ++adjacent_itr)
		{
			pcl::Supervoxel<pcl::PointXYZ>::Ptr neighbor_supervoxel = supervoxel_clusters.at(adjacent_itr->second);
			adjacent_supervoxel_centers.push_back(neighbor_supervoxel->centroid_);
		}
		std::stringstream ss;
		ss << "supervoxel_" << super_label;
		addSupervoxelConnectionsToViewer(super_cloud->centroid_, adjacent_supervoxel_centers, ss.str(), viewer1);
		label_itr = SupervoxelAdjacency.upper_bound(super_label);
	}
	// 等待直到可视化窗口关闭
	while (!viewer1->wasStopped())
	{
		viewer1->spinOnce(100);
		boost::this_thread::sleep(boost::posix_time::microseconds(1000));
	}

	//return 0;

	 //***********************************************************************
	//LCCP分割
	float concavity_tolerance_threshold = 10;
	float smoothness_threshold = 0.8;
	uint32_t min_segment_size = 0;
	bool use_extended_convexity = false;
	bool use_sanity_criterion = false;
	pcl::LCCPSegmentation<PointT> lccp;
	lccp.setConcavityToleranceThreshold(concavity_tolerance_threshold);//CC效验beta值
	lccp.setSmoothnessCheck(true, voxel_resolution, seed_resolution, smoothness_threshold);
	lccp.setKFactor(k_factor);               //CC效验的k邻点
	lccp.setInputSupervoxels(supervoxel_clusters, supervoxel_adjacency);
	lccp.setMinSegmentSize(min_segment_size);//最小分割尺寸
	lccp.segment();

	pcl::PointCloud<pcl::PointXYZL>::Ptr sv_labeled_cloud = super.getLabeledCloud();
	pcl::PointCloud<pcl::PointXYZL>::Ptr lccp_labeled_cloud = sv_labeled_cloud->makeShared();
	lccp.relabelCloud(*lccp_labeled_cloud);
	SuperVoxelAdjacencyList sv_adjacency_list;
	lccp.getSVAdjacencyList(sv_adjacency_list);

	pcl::visualization::PCLVisualizer::Ptr viewer2(new pcl::visualization::PCLVisualizer("LCCP超体素分割"));
	viewer2->setWindowName("LCCP超体素分割");
	viewer2->addPointCloud(lccp_labeled_cloud, "LCCP超体素分割");
	viewer2->setPointCloudRenderingProperties(pcl::visualization::PCL_VISUALIZER_POINT_SIZE, 1, "LCCP超体素分割");
	viewer2->setPointCloudRenderingProperties(pcl::visualization::PCL_VISUALIZER_OPACITY, 0.5, "LCCP超体素分割");
	// 等待直到可视化窗口关闭
	while (!viewer2->wasStopped())
	{
		viewer2->spinOnce(100);
		boost::this_thread::sleep(boost::posix_time::microseconds(1000));
	}

	return 0;

}

三、实验结果

原数据

去除地面后

超体聚类过分割

LCCP分割

四、总结

从实验结果来看，LCCP算法在相似物体场景分割方面有着较好的表现，对于颜色类似但棱角分明的物体可使用该算法。

五、相关链接

[1]PCL-低层次视觉-点云分割（超体聚类）
[2]PCL_使用LCCP进行点云分割