PCL学习：点云特征-PFH示例

代码说明：

显示mesh.pcd中点对应的PFH特征直方图；

CMakeLists.txt

cmake_minimum_required(VERSION 2.8 FATAL_ERROR)
project(pfh_feature)
find_package(PCL 1.7 REQUIRED)
include_directories(${PCL_INCLUDE_DIRS})
link_directories(${PCL_LIBRARY_DIRS})
add_definitions(${PCL_DEFINITIONS})
add_executable (pfh_feature pfh_feature.cpp)
target_link_libraries (pfh_feature ${PCL_LIBRARIES})

pfh_feature.cpp 

#include<pcl/visualization/pcl_plotter.h>
#include <pcl/io/pcd_io.h>
#include <pcl/filters/filter.h>
#include <pcl/filters/voxel_grid.h>
#include <pcl/console/parse.h>
#include <pcl/kdtree/kdtree_flann.h>
#include <pcl/filters/normal_space.h>
#include <pcl/common/eigen.h>
#include <pcl/features/normal_3d.h>
#include <pcl/features/pfh.h>
#include <pcl/visualization/histogram_visualizer.h>
#include <iostream>

using namespace std;
using namespace pcl::visualization;
using namespace pcl::console;

int
main (int argc, char * argv [])
{
	if(argc<2)
	{
		std::cout<<".exe source.pcd -r 0.005 -ds 5"<<endl;
		return 0;
	}
	float voxel_re=0.005,ds_N=5;
	parse_argument (argc, argv, "-r", voxel_re);// 设置点云分辨率
	parse_argument (argc, argv, "-ds", ds_N);// 设置半径

	//调节下采样的分辨率以保持数据处理的速度。
	// 下采样 
	pcl::PointCloud<pcl::PointXYZ>::Ptr cloud_src (new pcl::PointCloud<pcl::PointXYZ>); 
	pcl::io::loadPCDFile (argv[1], *cloud_src);	
	std::vector<int> indices1; 
	pcl::removeNaNFromPointCloud (*cloud_src, *cloud_src, indices1); 
	pcl::PointCloud<pcl::PointXYZ>::Ptr ds_src (new pcl::PointCloud<pcl::PointXYZ>); 
	pcl::VoxelGrid<pcl::PointXYZ> grid; 
	grid.setLeafSize (voxel_re, voxel_re, voxel_re); 
	grid.setInputCloud (cloud_src); 
	grid.filter (*ds_src); 

	//计算法向量
	pcl::PointCloud<pcl::Normal>::Ptr norm_src (new pcl::PointCloud<pcl::Normal>); 
	pcl::search::KdTree<pcl::PointXYZ>::Ptr tree_src(new pcl::search::KdTree<pcl::PointXYZ>()); 
	pcl::NormalEstimation<pcl::PointXYZ, pcl::Normal> ne; 
	PCL_INFO ("Normal Estimation - Source\n");	
	ne.setInputCloud (ds_src); 
	ne.setSearchSurface (cloud_src); 
	ne.setSearchMethod (tree_src); 
	ne.setRadiusSearch (ds_N*2*voxel_re); 
	ne.compute (*norm_src); 

	// 提取关键点
	pcl::PointCloud<pcl::PointXYZ>::Ptr keypoints_src (new pcl::PointCloud<pcl::PointXYZ>); 
	pcl::PointCloud<pcl::PointXYZ>::Ptr keypoints_tgt (new pcl::PointCloud<pcl::PointXYZ>); 
	grid.setLeafSize (ds_N*voxel_re,ds_N*voxel_re,ds_N*voxel_re); 
	grid.setInputCloud (ds_src); 
	grid.filter (*keypoints_src); 

	//Feature-Descriptor 
	PCL_INFO ("PFH - Feature Descriptor\n"); 
	//PFH	
	//PFH Source 
	pcl::PFHEstimation<pcl::PointXYZ, pcl::Normal, pcl::PFHSignature125> pfh_est_src; 
	pcl::search::KdTree<pcl::PointXYZ>::Ptr tree_pfh_src (new pcl::search::KdTree<pcl::PointXYZ>); 

	pfh_est_src.setSearchSurface (ds_src);//输入完整点云数据
	pfh_est_src.setInputCloud (keypoints_src); // 输入关键点
	pfh_est_src.setInputNormals (norm_src); 
	pfh_est_src.setRadiusSearch (2*ds_N*voxel_re);
	pfh_est_src.setSearchMethod(tree_pfh_src);
	pcl::PointCloud<pcl::PFHSignature125>::Ptr pfh_src (new pcl::PointCloud<pcl::PFHSignature125>);
	pfh_est_src.compute (*pfh_src); 

	//定义绘图器
	PCLPlotter *plotter = new PCLPlotter ("My Plotter"); 
	//设置特性
	plotter->setShowLegend (true);
	std::cout<<pcl::getFieldsList<pcl::PFHSignature125>(*pfh_src);
	
	//显示
	for (int m=0; m<6;m++)
	{
		plotter->addFeatureHistogram<pcl::PFHSignature125>(*pfh_src, "pfh", m, std::to_string(m)/*"one_fpfh"*/);
		plotter->setWindowSize(800, 600);
		plotter->spinOnce(1000);	
	}
	plotter->clearPlots();
    return 1;
}

cmake编译；

cmd命令：.\pfh_feature.exe ..\..\source\mesh.pcd

显示结果：