一、简介
在完成初始配准的情况下,一般采用ICP算法进行精细配准,标准的ICP配准算法是对源点云中的每个点,在目标点云中采用穷尽式凑所方法搜索距离最近的点作为对应点。然后求出配准对其所有对应点对的变换矩阵,最后将源点云数据通过该矩阵进行变换。算法一般设计一种或多种收敛规则,重复迭代地进行上述操作知道设计的收敛规则条件满足。如果不考虑测量误差,ICP算法的精度收到测量采样密度的影响,误差值正比于平均采样距离,即采样密度越高,拼合的精度也就越高。
二、代码分析
1)为了进行测试,首先创建一变换矩阵,对原始模型进行变换,然后再将变换后的模型与原始模型进行配准,以测试ICP:
// 旋转矩阵的具体定义 (请参考 https://en.wikipedia.org/wiki/Rotation_matrix)
double theta = M_PI / 20; // 设置旋转弧度的角度
transformation_matrix (0, 0) = cos (theta);
transformation_matrix (0, 1) = -sin (theta);
transformation_matrix (1, 0) = sin (theta);
transformation_matrix (1, 1) = cos (theta);
// 设置平移矩阵
transformation_matrix (0, 3) = 0.0;
transformation_matrix (1, 3) = 0.0;
transformation_matrix (2, 3) = 0.0;
// 在终端输出转换矩阵
std::cout << "Applying this rigid transformation to: cloud_in -> cloud_icp" << std::endl;
print4x4Matrix (transformation_matrix);
// 执行初始变换
pcl::transformPointCloud (*cloud_in, *cloud_icp, transformation_matrix);
2)利用刚体变换对初始点云进行转换,cloud_in为初始模型,cloud_icp为变换后的模型,cloud_tr是变换后数据的备份用于后期显示:
// 设置ICP配准算法输入参数
time.tic ();
pcl::IterativeClosestPoint<PointT, PointT> icp;
icp.setMaximumIterations (iterations);//设置迭代次数
icp.setInputSource (cloud_icp);
icp.setInputTarget (cloud_in);
icp.align (*cloud_icp);
3)创建ICP配准对象,设置ICP算法参数,setMaximumIterations函数设置ICP每次迭代最大次数为1,主要是为了后续对每一步迭代的效果进行可视化。然后 ,变换模型在初次配准进行后,当后续ICP配准再被调用时(当用户按下space按键时),每次调用迭代次数都与这里设置的参数一致:
if (icp.hasConverged ())
{
std::cout << "\nICP has converged, score is " << icp.getFitnessScore () << std::endl;
std::cout << "\nICP transformation " << iterations << " : cloud_icp -> cloud_in" << std::endl;
transformation_matrix = icp.getFinalTransformation ().cast<double>();
print4x4Matrix (transformation_matrix);
}
else
{
PCL_ERROR ("\nICP has not converged.\n");
return (-1);
}
4)为了可视化,在绘制窗口中创建两个视口,添加初始点云到创建的两个视口,并用相同的变量txt_gray_1v1赋予初始点云颜色。初始转换后的点云被设置为绿色,添加到左视口,ICP配准后的点云被设置为红色,添加到右视口。对每个视口添加文字描述,以便用户了解该是口相关信息,需要注意的是,因为输入的迭代次数变量iteration为整型,需要转换其格式为字符串:
// 可视化部分
pcl::visualization::PCLVisualizer viewer ("ICP demo");
// 创建两个独立的视口
int v1 (0);
int v2 (1);
viewer.createViewPort (0.0, 0.0, 0.5, 1.0, v1);
viewer.createViewPort (0.5, 0.0, 1.0, 1.0, v2);
// 我们将要使用的颜色
float bckgr_gray_level = 0.0; // 黑色
float txt_gray_lvl = 1.0 - bckgr_gray_level;
// 设置初始点云为白色
pcl::visualization::PointCloudColorHandlerCustom<PointT> cloud_in_color_h (cloud_in, (int) 255 * txt_gray_lvl, (int) 255 * txt_gray_lvl,
(int) 255 * txt_gray_lvl);//赋予显示点云的颜色
viewer.addPointCloud (cloud_in, cloud_in_color_h, "cloud_in_v1", v1);
viewer.addPointCloud (cloud_in, cloud_in_color_h, "cloud_in_v2", v2);
// 设置初始转换后的点云为绿色
pcl::visualization::PointCloudColorHandlerCustom<PointT> cloud_tr_color_h (cloud_tr, 20, 180, 20);
viewer.addPointCloud (cloud_tr, cloud_tr_color_h, "cloud_tr_v1", v1);
// 设置ICP配准后的点云为绿色
pcl::visualization::PointCloudColorHandlerCustom<PointT> cloud_icp_color_h (cloud_icp, 180, 20, 20);
viewer.addPointCloud (cloud_icp, cloud_icp_color_h, "cloud_icp_v2", v2);
// 在两个视口,分别添加文字描述
viewer.addText ("White: Original point cloud\nGreen: Matrix transformed point cloud", 10, 15, 16, txt_gray_lvl, txt_gray_lvl, txt_gray_lvl, "icp_info_1", v1);
viewer.addText ("White: Original point cloud\nRed: ICP aligned point cloud", 10, 15, 16, txt_gray_lvl, txt_gray_lvl, txt_gray_lvl, "icp_info_2", v2);
5)根据变量bckgr_gray_level,设置两个视口的背景颜色,用户只需要按下C按键,就可以获取相机参数。然后我像这个函数复制参数以保存相机位置、朝向和焦点。当查看显示器窗口时,任何时候只要按下按键,通过registerKeyboardCallback函数,我们来回调相关函数,以实现用户交互式的进行配准。
当用户按下键盘的任何键,keyboardEventOccurred将会被调用,这个函数将会检查这个按键是否是空格键。如果是空格键,全局变量布尔变量next_iteration被设置为真,允许可视化循环进入下面部分代码,ICP配准对象被调用来配准点云。前面我们已经对该ICP配准对象设置了源和目标点云,且设置循环时的最大迭代次数为1,但要注意的是这里用函数setInputSource设置源点云模型为cloud_icp,配准输出也是cloud_icp。也就是说每次迭代配准的源点云模型其实就是上次迭代配准的结果:
// 设置背景颜色
viewer.setBackgroundColor (bckgr_gray_level, bckgr_gray_level, bckgr_gray_level, v1);
viewer.setBackgroundColor (bckgr_gray_level, bckgr_gray_level, bckgr_gray_level, v2);
// 设置相机位置和方向
viewer.setCameraPosition (-3.68332, 2.94092, 5.71266, 0.289847, 0.921947, -0.256907, 0);
viewer.setSize (1280, 1024); // 设置可视化窗口的尺寸
// 通过键盘,调用回调函数
viewer.registerKeyboardCallback (&keyboardEventOccurred, (void*) NULL);
// 设置显示器
while (!viewer.wasStopped ())
{
viewer.spinOnce ();
// 用户按下空格键
if (next_iteration)
{
// 配准算法开始迭代
time.tic ();
icp.align (*cloud_icp);
std::cout << "Applied 1 ICP iteration in " << time.toc () << " ms" << std::endl;
if (icp.hasConverged ())
{
printf ("\033[11A");
printf ("\nICP has converged, score is %+.0e\n", icp.getFitnessScore ());
std::cout << "\nICP transformation " << ++iterations << " : cloud_icp -> cloud_in" << std::endl;
transformation_matrix *= icp.getFinalTransformation ().cast<double>(); //
print4x4Matrix (transformation_matrix); // 输出初始矩阵和当前矩阵的转换矩阵
ss.str ("");
ss << iterations;
std::string iterations_cnt = "ICP iterations = " + ss.str ();
viewer.updateText (iterations_cnt, 10, 60, 16, txt_gray_lvl, txt_gray_lvl, txt_gray_lvl, "iterations_cnt");
viewer.updatePointCloud (cloud_icp, cloud_icp_color_h, "cloud_icp_v2");
}
else
{
PCL_ERROR ("\nICP has not converged.\n");
return (-1);
}
}
next_iteration = false;
}
6)整体代码:
#include <iostream>
#include <string>
#include <pcl/io/ply_io.h>
#include <pcl/point_types.h>
#include <pcl/registration/icp.h>
#include <pcl/visualization/pcl_visualizer.h>
#include <pcl/console/time.h>
typedef pcl::PointXYZ PointT; //定义全局变量
typedef pcl::PointCloud<PointT> PointCloudT;
bool next_iteration = false;
void print4x4Matrix(const Eigen::Matrix4d& matrix) //创建一个矩阵输出的函数
{
printf("Rotation matrix :\n");
printf(" |%6.3f %6.3f %6.3f|\n", matrix(0, 0), matrix(0, 1), matrix(0, 2));
printf(" |%6.3f %6.3f %6.3f|\n", matrix(1, 0), matrix(1, 1), matrix(1, 2));
printf(" |%6.3f %6.3f %6.3f|\n", matrix(2, 0), matrix(2, 1), matrix(2, 2));
printf("Translation vector :\n");
printf(" |%6.3f %6.3f %6.3f|\n", matrix(0, 3), matrix(1, 3), matrix(2, 3));
}
void keyboardEventOccurred(const pcl::visualization::KeyboardEvent& event, void* nothing) //创建键盘回调函数,按下以后更改是否迭代的标志位
{
if (event.getKeySym() == "space" && event.keyDown())
next_iteration = true;
}
int main(int argc, char* argv[])
{
PointCloudT::Ptr cloud_in(new PointCloudT);
PointCloudT::Ptr cloud_tr(new PointCloudT);
PointCloudT::Ptr cloud_icp(new PointCloudT);
if (argc < 2) //初始化输入的迭代次数
{
printf("Usage:\n");
printf("\t\t%s file.ply number_of_ICP_iterations\n", argv[0]);
PCL_ERROR("Provide one ply file.\n");
return(-1);
}
int iterations = 1;
if (argc > 2)
{
iterations = atoi(argv[2]);
if (iterations < 1)
{
PCL_ERROR("Number of initial iterations must be >= 1\n");
return(-1);
}
}
pcl::console::TicToc time; //实例化一个计时器
time.tic();
if (pcl::io::loadPLYFile(argv[1], *cloud_in) < 0) //加载点云并输出所使用的时间
{
PCL_ERROR("Error loading cloud %s.\n", argv[1]);
return (-1);
}
std::cout << "\nLoaded file " << argv[1] << "(" << cloud_in->size() << "points)in " << time.toc() << "ms\n" << std::endl;
Eigen::Matrix4d transformation_matrix = Eigen::Matrix4d::Identity(); //实例化一个旋转矩阵
double theta = M_PI / 20; //设置旋转矩阵的旋转角度
transformation_matrix(0, 0) = cos(theta); //初始化旋转矩阵
transformation_matrix(0, 1) = -sin(theta);
transformation_matrix(1, 0) = sin(theta);
transformation_matrix(1, 1) = cos(theta);
transformation_matrix(0, 3) = 0.0; //设置平移矩阵
transformation_matrix(1, 3) = 0.0;
transformation_matrix(2, 3) = 0.0;
std::cout << "Applying this rigid transformation to: cloud_in -> cloud_icp" << std::endl;
print4x4Matrix(transformation_matrix);
pcl::transformPointCloud(*cloud_in, *cloud_icp, transformation_matrix); //执行初始变换,cloud_in为初始模型,cloud_icp为变换后的模型
*cloud_tr = *cloud_icp; //cloud_tr是变换后的数据的备份用于后期的显示
time.tic();
pcl::IterativeClosestPoint<PointT, PointT>icp; //实例化一个icp对象
icp.setMaximumIterations(iterations); //设置迭代次数
icp.setInputSource(cloud_icp);
icp.setInputTarget(cloud_in);
icp.align(*cloud_icp);
icp.setMaximumIterations(1);
std::cout << "Applied" << iterations << "ICP iterations in " << time.toc() << "ms" << std::endl;
if (icp.hasConverged()) //当或许ICP配准再被调用时,每次调用迭代次数都与这里的参数一致
{
std::cout << "\nICP has converged, score is " << icp.getFitnessScore() << std::endl;
std::cout << "\nICP transformation " << iterations << " :cloud_icp -> cloud_in" << std::endl;
transformation_matrix = icp.getFinalTransformation().cast<double>();
print4x4Matrix(transformation_matrix);
}
else
{
PCL_ERROR("\nICP has not converged.\n");
return(-1);
}
pcl::visualization::PCLVisualizer viewer("ICP demo"); //初始化两个视口用于显示配准前后的点云
int v1(0);
int v2(1);
viewer.createViewPort(0.0, 0.0, 0.5, 1.0, v1);
viewer.createViewPort(0.5, 0.0, 1.0, 1.0, v2);
float bckgr_gray_level = 0.0; //使用到的颜色为黑色
float txt_gray_1v1 = 1.0 - bckgr_gray_level;
pcl::visualization::PointCloudColorHandlerCustom<PointT> cloud_in_color_h(cloud_in, (int)255 * txt_gray_1v1, (int)255 * txt_gray_1v1, (int)255 * txt_gray_1v1);
viewer.addPointCloud(cloud_in, cloud_in_color_h, "cloud_in_v1", v1);
viewer.addPointCloud(cloud_in, cloud_in_color_h, "cloud_in_v2", v2);
pcl::visualization::PointCloudColorHandlerCustom<PointT> cloud_tr_color_h(cloud_tr, 20, 180, 20);
viewer.addPointCloud(cloud_tr, cloud_tr_color_h, "cloud_tr_v1", v1); //设置初始转换后的颜色
pcl::visualization::PointCloudColorHandlerCustom<PointT> cloud_icp_color_h(cloud_icp, 180, 20, 20);
viewer.addPointCloud(cloud_icp, cloud_icp_color_h, "cloud_icp_v2", v2); //设置icp配准后的颜色
viewer.addText("White: original point cloud\nGreen: Matrix transformed point cloud", 10, 15, 16, txt_gray_1v1, txt_gray_1v1, txt_gray_1v1, "icp_info_1", v1);
viewer.addText("White: original point cloud\nRed: ICP aligned point cloud", 10, 15, 16, txt_gray_1v1, txt_gray_1v1, txt_gray_1v1, "icp_info_2", v2);
std::stringstream ss; //输出进行的迭代次数
ss << iterations;
std::string iterations_cnt = "ICP iterations = " + ss.str();
viewer.addText(iterations_cnt, 10, 60, 16, txt_gray_1v1, txt_gray_1v1, txt_gray_1v1, "iterations_cnt", v2);
viewer.setBackgroundColor(bckgr_gray_level, bckgr_gray_level, bckgr_gray_level, v1); //设置视口的背景颜色
viewer.setBackgroundColor(bckgr_gray_level, bckgr_gray_level, bckgr_gray_level, v2);
viewer.setCameraPosition(-3.68332, 2.94092, 5.71266, 0.289847, 0.921947, -0.256907, 0); //设置相机位置和方向
viewer.setSize(1280, 1024); //设置可视化窗口的尺寸
viewer.registerKeyboardCallback(&keyboardEventOccurred, (void*)NULL); //通过键盘调用回调函数
while (!viewer.wasStopped()) //设置显示窗口
{
viewer.spinOnce();
if (next_iteration) //开始进行icp配准
{
time.tic();
icp.align(*cloud_icp);
std::cout << "Applied 1 ICP iteration in " << time.toc() << "ms" << std::endl;
if (icp.hasConverged())
{
printf("\033[11A");
printf("\nICP has converged, score is %+.0e\n", icp.getFitnessScore());
std::cout << "\nICP transformation " << ++iterations << " :cloud_icp -> cloud_in" << std::endl;
transformation_matrix *= icp.getFinalTransformation().cast<double>();
print4x4Matrix(transformation_matrix); //输出初始矩阵和当前矩阵的转换矩阵
ss.str("");
ss << iterations;
std::string iterations_cnt = "ICP iterations = " + ss.str();
viewer.updateText(iterations_cnt, 10, 60, 16, txt_gray_1v1, txt_gray_1v1, txt_gray_1v1, "iterations_cnt");
viewer.updatePointCloud(cloud_icp, cloud_icp_color_h, "cloud_icp_v2");
}
else
{
PCL_ERROR("\nICP has not converged.\n");
return(-1);
}
}
next_iteration = false;
}
return(0);
}
三、编译结果
1)迭代次数为1次时:
2)迭代次数为8次时:
3)迭代次数为20次时: