两点云求差集和交集

给算法爸爸上香

已于 2023-05-11 21:33:34 修改

阅读量2.5k

点赞数 6

CC 4.0 BY-SA版权

分类专栏： PCL 文章标签：点云 PCL

于 2022-11-13 17:34:41 首次发布

本文链接：https://blog.youkuaiyun.com/taifyang/article/details/127834570

PCL 专栏收录该内容

26 篇文章

订阅专栏

本文介绍了如何使用kd-tree数据结构高效地计算两个点云的差集和交集。在点云2上构建kd-tree，通过搜索半径找到点云1中与点云2接近的点，然后根据搜索结果生成差集或交集。代码示例展示了在PCL库中实现这一过程，并保存结果到新的点云文件。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

这里两点云的差集指从点云1中删除属于点云2的点得到的点集，交集指既属于点云1又属于点云2的点集。

两点云求差集

基于kd-tree搜索的方法较快速，当然也可以暴力搜索。思路如下：
step1 在点云2建立kd-tree，设置容忍误差（搜索半径）
step2 遍历点云1中的点，记录下到点云2中的点的距离小于搜索半径的点的索引
step3 方法一：将点云1中不在索引中的点保存下来作为结果点云3
方法二：直接对点云1操作，删除索引点
建议使用方法一，代码实现更简单，而且速度较快（erase操作会使迭代器位置改变，影响效率）；除非要对原点云操作用方法二。

#include <iostream>
#include <pcl/io/pcd_io.h>
#include <pcl/point_types.h>
#include <pcl/common/io.h>
#include <pcl/search/kdtree.h> 


int main(int argc, char** argv)
{
	pcl::PointCloud<pcl::PointXYZ>::Ptr cloud1(new pcl::PointCloud<pcl::PointXYZ>);
	pcl::PointCloud<pcl::PointXYZ>::Ptr cloud2(new pcl::PointCloud<pcl::PointXYZ>);

	pcl::io::loadPCDFile("bunny1.pcd", *cloud1);
	pcl::io::loadPCDFile("bunny2.pcd", *cloud2);

	pcl::search::KdTree<pcl::PointXYZ>::Ptr kdtree(new pcl::search::KdTree<pcl::PointXYZ>);
	kdtree->setInputCloud(cloud2);

	std::vector<int> pointIdxRadiusSearch;
	std::vector<float> pointRadiusSquaredDistance;
	std::vector<int> indices;
	float radius = 0.0001;

	for (size_t i = 0; i < cloud1->size(); i++)
	{
		if (kdtree->radiusSearch(cloud1->points[i], radius, pointIdxRadiusSearch, pointRadiusSquaredDistance) > 0)
		{
			indices.push_back(i);
		}
	}

	pcl::PointCloud<pcl::PointXYZ>::Ptr cloud3(new pcl::PointCloud<pcl::PointXYZ>);
	for (size_t i = 0; i < cloud1->size(); i++)
	{
		if (find(indices.begin(), indices.end(), i) == indices.end())
			cloud3->push_back(cloud1->points[i]);
	}

	//for (size_t i = 0; i < indices.size(); i++)
	//{
	//	cloud1->erase(cloud1->begin() + indices[i] - i);
	//}

	std::cout << "cloud1 has " << cloud1->size() << " points" << std::endl;
	std::cout << "cloud2 has " << cloud2->size() << " points" << std::endl;
	std::cout << "cloud3 has " << cloud3->size() << " points" << std::endl;
	pcl::io::savePCDFile("cloud3.pcd", *cloud3);

	system("pause");
	return 0;
}

两点云求交集

#include <iostream>
#include <pcl/io/pcd_io.h>
#include <pcl/point_types.h>
#include <pcl/common/io.h>
#include <pcl/search/kdtree.h> 


int main(int argc, char** argv)
{
	pcl::PointCloud<pcl::PointXYZ>::Ptr cloud1(new pcl::PointCloud<pcl::PointXYZ>);
	pcl::PointCloud<pcl::PointXYZ>::Ptr cloud2(new pcl::PointCloud<pcl::PointXYZ>);

	pcl::io::loadPCDFile("bunny1.pcd", *cloud1);
	pcl::io::loadPCDFile("bunny2.pcd", *cloud2);

	pcl::search::KdTree<pcl::PointXYZ>::Ptr kdtree(new pcl::search::KdTree<pcl::PointXYZ>);
	kdtree->setInputCloud(cloud2);

	std::vector<int> pointIdxRadiusSearch;
	std::vector<float> pointRadiusSquaredDistance;
	std::vector<int> indices;
	float radius = 0.0001;

	for (size_t i = 0; i < cloud1->size(); i++)
	{
		if (kdtree->radiusSearch(cloud1->points[i], radius, pointIdxRadiusSearch, pointRadiusSquaredDistance) > 0)
		{
			indices.push_back(i);
		}
	}

	pcl::PointCloud<pcl::PointXYZ>::Ptr cloud3(new pcl::PointCloud<pcl::PointXYZ>);
	for (size_t i = 0; i < cloud1->size(); i++)
	{
		if (find(indices.begin(), indices.end(), i) != indices.end())
			cloud3->push_back(cloud1->points[i]);
	}

	std::cout << "cloud1 has " << cloud1->size() << " points" << std::endl;
	std::cout << "cloud2 has " << cloud2->size() << " points" << std::endl;
	std::cout << "cloud3 has " << cloud3->size() << " points" << std::endl;
	pcl::io::savePCDFile("cloud3.pcd", *cloud3);

	system("pause");
	return 0;
}