浅析ORB、SURF、SIFT特征点提取方法以及ICP匹配方法
程序员文章站
2021-12-14 10:42:05
目录main.cppcmakelists.txt执行效果icpcmakelists.txt执行效果在进行编译视觉slam时，书中提到了orb、surf、sift提取方法，以及特征提取方法暴力匹配(br...
在进行编译视觉slam时，书中提到了orb、surf、sift提取方法，以及特征提取方法暴力匹配(brute-force matcher)和快速近邻匹配(flann)。以及7.9讲述的3d-3d：迭代最近点(iterative closest point,icp)方法，icp 的求解方式有两种：利用线性代数求解(主要是svd)，以及利用非线性优化方式求解。
完整代码代码如下：
链接：https://pan.baidu.com/s/1rlh9jtg_awtuyzmphqij3q 提取码：8888
main.cpp
#include <iostream>
#include "opencv2/opencv.hpp"
#include "opencv2/core/core.hpp"
#include "opencv2/features2d/features2d.hpp"
#include "opencv2/highgui/highgui.hpp"
#include <opencv2/xfeatures2d.hpp>
#include <iostream>
#include <vector>
#include <time.h>
#include <chrono>
#include <math.h>
#include<bits/stdc++.h>
using namespace std;
using namespace cv;
using namespace cv::xfeatures2d;
double picture1_size_change=1;
double picture2_size_change=1;
bool show_picture = true;
void extract_orb2(string picture1, string picture2)
{
//-- 读取图像
mat img_1 = imread(picture1, cv_load_image_color);
mat img_2 = imread(picture2, cv_load_image_color);
assert(img_1.data != nullptr && img_2.data != nullptr);
resize(img_1, img_1, size(), picture1_size_change, picture1_size_change);
resize(img_2, img_2, size(), picture2_size_change, picture2_size_change);
//-- 初始化
std::vector<keypoint> keypoints_1, keypoints_2;
mat descriptors_1, descriptors_2;
ptr<featuredetector> detector = orb::create(2000,(1.200000048f), 8, 100);
ptr<descriptorextractor> descriptor = orb::create(5000);
ptr<descriptormatcher> matcher = descriptormatcher::create("bruteforce-hamming");
//-- 第一步:检测 oriented fast 角点位置
chrono::steady_clock::time_point t1 = chrono::steady_clock::now();
detector->detect(img_1, keypoints_1);
detector->detect(img_2, keypoints_2);
//-- 第二步:根据角点位置计算 brief 描述子
descriptor->compute(img_1, keypoints_1, descriptors_1);
descriptor->compute(img_2, keypoints_2, descriptors_2);
chrono::steady_clock::time_point t2 = chrono::steady_clock::now();
chrono::duration<double> time_used = chrono::duration_cast<chrono::duration<double>>(t2 - t1);
// cout << "extract orb cost = " << time_used.count() * 1000 << " ms " << endl;
cout << "detect " << keypoints_1.size() << " and " << keypoints_2.size() << " keypoints " << endl;
if (show_picture)
{
mat outimg1;
drawkeypoints(img_1, keypoints_1, outimg1, scalar::all(-1), drawmatchesflags::default);
imshow("orb features", outimg1);
}
//-- 第三步:对两幅图像中的brief描述子进行匹配，使用 hamming 距离
vector<dmatch> matches;
// t1 = chrono::steady_clock::now();
matcher->match(descriptors_1, descriptors_2, matches);
t2 = chrono::steady_clock::now();
time_used = chrono::duration_cast<chrono::duration<double>>(t2 - t1);
cout << "extract and match orb cost = " << time_used.count() * 1000 << " ms " << endl;
//-- 第四步:匹配点对筛选
// 计算最小距离和最大距离
auto min_max = minmax_element(matches.begin(), matches.end(),
[](const dmatch &m1, const dmatch &m2)
{ return m1.distance < m2.distance; });
double min_dist = min_max.first->distance;
double max_dist = min_max.second->distance;
// printf("-- max dist : %f \n", max_dist);
// printf("-- min dist : %f \n", min_dist);
//当描述子之间的距离大于两倍的最小距离时,即认为匹配有误.但有时候最小距离会非常小,设置一个经验值30作为下限.
std::vector<dmatch> good_matches;
for (int i = 0; i < descriptors_1.rows; i++)
{
if (matches[i].distance <= max(2 * min_dist, 30.0))
{
good_matches.push_back(matches[i]);
}
}
cout << "match " << good_matches.size() << " keypoints " << endl;
//-- 第五步:绘制匹配结果
mat img_match;
mat img_goodmatch;
drawmatches(img_1, keypoints_1, img_2, keypoints_2, matches, img_match);
drawmatches(img_1, keypoints_1, img_2, keypoints_2, good_matches, img_goodmatch);
if (show_picture)
imshow("good matches", img_goodmatch);
if (show_picture)
waitkey(0);
}
void extract_sift(string picture1, string picture2)
{
// double t = (double)gettickcount();
mat temp = imread(picture1, imread_grayscale);
mat image_check_changed = imread(picture2, imread_grayscale);
if (!temp.data || !image_check_changed.data)
{
printf("could not load images...\n");
return;
}
resize(temp, temp, size(), picture1_size_change, picture1_size_change);
resize(image_check_changed, image_check_changed, size(), picture2_size_change, picture2_size_change);
//mat image_check_changed = change_image(image_check);
//("temp", temp);
if (show_picture)
imshow("image_check_changed", image_check_changed);
int minhessian = 500;
// ptr<surf> detector = surf::create(minhessian); // surf
ptr<sift> detector = sift::create(); // sift
vector<keypoint> keypoints_obj;
vector<keypoint> keypoints_scene;
mat descriptor_obj, descriptor_scene;
clock_t starttime, endtime;
starttime = clock();
chrono::steady_clock::time_point t1 = chrono::steady_clock::now();
// cout << "extract orb cost = " << time_used.count() * 1000 << " ms " << endl;
detector->detectandcompute(temp, mat(), keypoints_obj, descriptor_obj);
detector->detectandcompute(image_check_changed, mat(), keypoints_scene, descriptor_scene);
cout << "detect " << keypoints_obj.size() << " and " << keypoints_scene.size() << " keypoints " << endl;
// matching
flannbasedmatcher matcher;
vector<dmatch> matches;
matcher.match(descriptor_obj, descriptor_scene, matches);
chrono::steady_clock::time_point t2 = chrono::steady_clock::now();
chrono::duration<double> time_used = chrono::duration_cast<chrono::duration<double>>(t2 - t1);
cout << "extract and match cost = " << time_used.count() * 1000 << " ms " << endl;
//求最小最大距离
double mindist = 1000;
double maxdist = 0;
//row--行 col--列
for (int i = 0; i < descriptor_obj.rows; i++)
{
double dist = matches[i].distance;
if (dist > maxdist)
{
maxdist = dist;
}
if (dist < mindist)
{
mindist = dist;
}
}
// printf("max distance : %f\n", maxdist);
// printf("min distance : %f\n", mindist);
// find good matched points
vector<dmatch> goodmatches;
for (int i = 0; i < descriptor_obj.rows; i++)
{
double dist = matches[i].distance;
if (dist < max(5 * mindist, 1.0))
{
goodmatches.push_back(matches[i]);
}
}
//rectangle(temp, point(1, 1), point(177, 157), scalar(0, 0, 255), 8, 0);
cout << "match " << goodmatches.size() << " keypoints " << endl;
endtime = clock();
// cout << "took time : " << (double)(endtime - starttime) / clocks_per_sec * 1000 << " ms" << endl;
mat matchesimg;
drawmatches(temp, keypoints_obj, image_check_changed, keypoints_scene, goodmatches, matchesimg, scalar::all(-1),
scalar::all(-1), vector<char>(), drawmatchesflags::not_draw_single_points);
if (show_picture)
imshow("flann matching result01", matchesimg);
// imwrite("c:/users/administrator/desktop/matchesimg04.jpg", matchesimg);
//求h
std::vector<point2f> points1, points2;
//保存对应点
for (size_t i = 0; i < goodmatches.size(); i++)
{
//queryidx是对齐图像的描述子和特征点的下标。
points1.push_back(keypoints_obj[goodmatches[i].queryidx].pt);
//queryidx是是样本图像的描述子和特征点的下标。
points2.push_back(keypoints_scene[goodmatches[i].trainidx].pt);
}
// find homography 计算homography，ransac随机抽样一致性算法
mat h = findhomography(points1, points2, ransac);
//imwrite("c:/users/administrator/desktop/c-train/c-train/result/sift/image4_surf_minhessian1000_ mindist1000_a0.9b70.jpg", matchesimg);
vector<point2f> obj_corners(4);
vector<point2f> scene_corners(4);
obj_corners[0] = point(0, 0);
obj_corners[1] = point(temp.cols, 0);
obj_corners[2] = point(temp.cols, temp.rows);
obj_corners[3] = point(0, temp.rows);
//透视变换(把斜的图片扶正)
perspectivetransform(obj_corners, scene_corners, h);
//mat dst;
cvtcolor(image_check_changed, image_check_changed, color_gray2bgr);
line(image_check_changed, scene_corners[0], scene_corners[1], scalar(0, 0, 255), 2, 8, 0);
line(image_check_changed, scene_corners[1], scene_corners[2], scalar(0, 0, 255), 2, 8, 0);
line(image_check_changed, scene_corners[2], scene_corners[3], scalar(0, 0, 255), 2, 8, 0);
line(image_check_changed, scene_corners[3], scene_corners[0], scalar(0, 0, 255), 2, 8, 0);
if (show_picture)
{
mat outimg1;
mat temp_color = imread(picture1, cv_load_image_color);
drawkeypoints(temp_color, keypoints_obj, outimg1, scalar::all(-1), drawmatchesflags::default);
imshow("sift features", outimg1);
}
if (show_picture)
imshow("draw object", image_check_changed);
// imwrite("c:/users/administrator/desktop/image04.jpg", image_check_changed);
// t = ((double)gettickcount() - t) / gettickfrequency();
// printf("averagetime:%f\n", t);
if (show_picture)
waitkey(0);
}
void extract_surf(string picture1, string picture2)
{
// double t = (double)gettickcount();
mat temp = imread(picture1, imread_grayscale);
mat image_check_changed = imread(picture2, imread_grayscale);
if (!temp.data || !image_check_changed.data)
{
printf("could not load images...\n");
return;
}
resize(temp, temp, size(), picture1_size_change, picture1_size_change);
resize(image_check_changed, image_check_changed, size(), picture2_size_change, picture2_size_change);
//mat image_check_changed = change_image(image_check);
//("temp", temp);
if (show_picture)
imshow("image_check_changed", image_check_changed);
int minhessian = 500;
ptr<surf> detector = surf::create(minhessian); // surf
// ptr<sift> detector = sift::create(minhessian); // sift
vector<keypoint> keypoints_obj;
vector<keypoint> keypoints_scene;
mat descriptor_obj, descriptor_scene;
clock_t starttime, endtime;
starttime = clock();
chrono::steady_clock::time_point t1 = chrono::steady_clock::now();
// cout << "extract orb cost = " << time_used.count() * 1000 << " ms " << endl;
detector->detectandcompute(temp, mat(), keypoints_obj, descriptor_obj);
detector->detectandcompute(image_check_changed, mat(), keypoints_scene, descriptor_scene);
cout << "detect " << keypoints_obj.size() << " and " << keypoints_scene.size() << " keypoints " << endl;
// matching
flannbasedmatcher matcher;
vector<dmatch> matches;
matcher.match(descriptor_obj, descriptor_scene, matches);
chrono::steady_clock::time_point t2 = chrono::steady_clock::now();
chrono::duration<double> time_used = chrono::duration_cast<chrono::duration<double>>(t2 - t1);
cout << "extract and match cost = " << time_used.count() * 1000 << " ms " << endl;
//求最小最大距离
double mindist = 1000;
double maxdist = 0;
//row--行 col--列
for (int i = 0; i < descriptor_obj.rows; i++)
{
double dist = matches[i].distance;
if (dist > maxdist)
{
maxdist = dist;
}
if (dist < mindist)
{
mindist = dist;
}
}
// printf("max distance : %f\n", maxdist);
// printf("min distance : %f\n", mindist);
// find good matched points
vector<dmatch> goodmatches;
for (int i = 0; i < descriptor_obj.rows; i++)
{
double dist = matches[i].distance;
if (dist < max(2 * mindist, 0.15))
{
goodmatches.push_back(matches[i]);
}
}
//rectangle(temp, point(1, 1), point(177, 157), scalar(0, 0, 255), 8, 0);
cout << "match " << goodmatches.size() << " keypoints " << endl;
endtime = clock();
// cout << "took time : " << (double)(endtime - starttime) / clocks_per_sec * 1000 << " ms" << endl;
mat matchesimg;
drawmatches(temp, keypoints_obj, image_check_changed, keypoints_scene, goodmatches, matchesimg, scalar::all(-1),
scalar::all(-1), vector<char>(), drawmatchesflags::not_draw_single_points);
if (show_picture)
imshow("flann matching result01", matchesimg);
// imwrite("c:/users/administrator/desktop/matchesimg04.jpg", matchesimg);
//求h
std::vector<point2f> points1, points2;
//保存对应点
for (size_t i = 0; i < goodmatches.size(); i++)
{
//queryidx是对齐图像的描述子和特征点的下标。
points1.push_back(keypoints_obj[goodmatches[i].queryidx].pt);
//queryidx是是样本图像的描述子和特征点的下标。
points2.push_back(keypoints_scene[goodmatches[i].trainidx].pt);
}
// find homography 计算homography，ransac随机抽样一致性算法
mat h = findhomography(points1, points2, ransac);
//imwrite("c:/users/administrator/desktop/c-train/c-train/result/sift/image4_surf_minhessian1000_ mindist1000_a0.9b70.jpg", matchesimg);
vector<point2f> obj_corners(4);
vector<point2f> scene_corners(4);
obj_corners[0] = point(0, 0);
obj_corners[1] = point(temp.cols, 0);
obj_corners[2] = point(temp.cols, temp.rows);
obj_corners[3] = point(0, temp.rows);
//透视变换(把斜的图片扶正)
perspectivetransform(obj_corners, scene_corners, h);
//mat dst;
cvtcolor(image_check_changed, image_check_changed, color_gray2bgr);
line(image_check_changed, scene_corners[0], scene_corners[1], scalar(0, 0, 255), 2, 8, 0);
line(image_check_changed, scene_corners[1], scene_corners[2], scalar(0, 0, 255), 2, 8, 0);
line(image_check_changed, scene_corners[2], scene_corners[3], scalar(0, 0, 255), 2, 8, 0);
line(image_check_changed, scene_corners[3], scene_corners[0], scalar(0, 0, 255), 2, 8, 0);
if (show_picture)
{
mat outimg1;
mat temp_color = imread(picture1, cv_load_image_color);
drawkeypoints(temp_color, keypoints_obj, outimg1, scalar::all(-1), drawmatchesflags::default);
imshow("surf features", outimg1);
}
if (show_picture)
imshow("draw object", image_check_changed);
// imwrite("c:/users/administrator/desktop/image04.jpg", image_check_changed);
// t = ((double)gettickcount() - t) / gettickfrequency();
// printf("averagetime:%f\n", t);
if (show_picture)
waitkey(0);
}
void extract_akaze(string picture1,string picture2)
{
//读取图片
mat temp = imread(picture1,imread_grayscale);
mat image_check_changed = imread(picture2,imread_grayscale);
//如果不能读到其中任何一张图片，则打印不能下载图片
if(!temp.data || !image_check_changed.data)
{
printf("could not load iamges...\n");
return;
}
resize(temp,temp,size(),picture1_size_change,picture1_size_change);
resize(image_check_changed,image_check_changed,size(),picture2_size_change,picture2_size_change);
//mat image_check_changed = change_image(image_check);
//("temp", temp);
if(show_picture)
{
imshow("image_checked_changed",image_check_changed);
}
int minhessian=500;
ptr<akaze> detector=akaze::create();//akaze
vector<keypoint> keypoints_obj;
vector<keypoint> keypoints_scene;
mat descriptor_obj,descriptor_scene;
clock_t starttime,endtime;
starttime=clock();
chrono::steady_clock::time_point t1=chrono::steady_clock::now();
detector->detectandcompute(temp,mat(),keypoints_obj,descriptor_obj);
detector->detectandcompute(image_check_changed,mat(),keypoints_scene,descriptor_scene);
cout<<" detect "<<keypoints_obj.size()<<" and "<<keypoints_scene.size<<" keypoints "<<endl;
//matching
flannbasedmatcher matcher;
vector<dmatch> matches;
matcher.match(descriptor_obj,descriptor_scene,matches);
chrono::steady_clock::time_point t2 = chrono::steady_clock::now();
chrono::duration<double> time_used = chrono::duration_cast<chrono::duration<double>>(t2-t1);
cout << "extract and match cost = " << time_used.count()*1000<<" ms "<<endl;
//求最小最大距离
double mindist = 1000;
double max_dist = 0;
//row--行 col--列
for(int i=0;i<descriptor_obj.rows;i++)
{
double dist = match[i].distance;
if(dist > maxdist)
{
maxdist = dist;
}
if(dist<mindist)
{
mindist = dist;
}
}
// printf("max distance : %f\n", maxdist);
// printf("min distance : %f\n", mindist);
// find good matched points
vector<dmatch> goodmatches;
for(imt i=0;i<descriptor_obj.rows;i++)
{
double dist = matches[i].distance;
if(dist < max(5 * mindist,1.0))
{
goodmatches.push_back(matches[i]);
}
}
//rectangle(temp, point(1, 1), point(177, 157), scalar(0, 0, 255), 8, 0);
cout<<" match "<<goodmatches.size()<<" keypoints "<<endl;
endtime = clock();
// cout << "took time : " << (double)(endtime - starttime) / clocks_per_sec * 1000 << " ms" << endl;
mat matchesimg;
drawmatches(temp,keypoints_obj,image_check_changed,keypoints_scene,goodmatches,
matchesimg,scalar::all(-1),
scalar::all(-1),vector<char>(),drawmatchesflags::not_draw_single_points);
if(show_picture)
imshow("flann matching result01",matchesimg);
// imwrite("c:/users/administrator/desktop/matchesimg04.jpg", matchesimg);
//求h
std::vector<point2f> points1,points2;
//保存对应点
for(size_t i = 0;i < goodmatches.size();i++)
{
//queryidx是对齐图像的描述子和特征点的下标。
points1.push_back(keypoints_obj[goodmatches[i].queryidx].pt);
//queryidx是是样本图像的描述子和特征点的下标。
points2.push_back(keypoints_scene[goodmatches[i].trainidx].pt);
}
// find homography 计算homography，ransac随机抽样一致性算法
mat h = findhomography(points1,points2,ransac);
//imwrite("c:/users/administrator/desktop/c-train/c-train/result/sift/image4_surf_minhessian1000_ mindist1000_a0.9b70.jpg", matchesimg);
vector<point2f> obj_corners(4);
vector<point2f> scene_corners(4);
obj_corners[0] = point(0,0);
obj_corners[0] = point(temp.count,0);
obj_corners[0] = point(temp.cols,temp.rows);
obj_corners[0] = point(0,temp.rows);
//透视变换(把斜的图片扶正)
perspectivetransform(obj_corners,scene_corners,h);
//mat dst
cvtcolor(image_check_changed,image_check_changed,color_gray2bgr);
line(image_check_changed,scene_corners[0],scene_corners[1],scalar(0,0,255),2,8,0);
line(image_check_changed,scene_corners[1],scene_corners[2],scalar(0,0,255),2,8,0);
line(image_check_changed,scene_corners[2],scene_corners[3],scalar(0,0,255),2,8,0);
line(image_check_changed,scene_corners[3],scene_corners[0],scalar(0,0,255),2,8,0);
if(show_picture)
{
mat outimg1;
mat temp_color = imread(picture1,cv_load_image_color);
drawkeypoints(temp_color,keypoints_obj,outimg1,scalar::all(-1),drawmatchesflags::default);
imshow("akaze features",outimg1);
}
if(show_picture)
waitkey(0);
}
void extract_orb(string picture1, string picture2)
{
mat img_1 = imread(picture1);
mat img_2 = imread(picture2);
resize(img_1, img_1, size(), picture1_size_change, picture1_size_change);
resize(img_2, img_2, size(), picture2_size_change, picture2_size_change);
if (!img_1.data || !img_2.data)
{
cout << "error reading images " << endl;
return ;
}
vector<point2f> recognized;
vector<point2f> scene;
recognized.resize(1000);
scene.resize(1000);
mat d_srcl, d_srcr;
mat img_matches, des_l, des_r;
//orb算法的目标必须是灰度图像
cvtcolor(img_1, d_srcl, color_bgr2gray);//cpu版的orb算法源码中自带对输入图像灰度化，此步可省略
cvtcolor(img_2, d_srcr, color_bgr2gray);
ptr<orb> d_orb = orb::create(1500);
mat d_descriptorsl, d_descriptorsr, d_descriptorsl_32f, d_descriptorsr_32f;
vector<keypoint> keypoints_1, keypoints_2;
//设置关键点间的匹配方式为norm_l2，更建议使用 flannbased = 1, bruteforce = 2, bruteforce_l1 = 3, bruteforce_hamming = 4, bruteforce_hamminglut = 5, bruteforce_sl2 = 6
ptr<descriptormatcher> d_matcher = descriptormatcher::create(norm_l2);
std::vector<dmatch> matches;//普通匹配
std::vector<dmatch> good_matches;//通过keypoint之间距离筛选匹配度高的匹配结果
clock_t starttime, endtime;
starttime = clock();
chrono::steady_clock::time_point t1 = chrono::steady_clock::now();
d_orb -> detectandcompute(d_srcl, mat(), keypoints_1, d_descriptorsl);
d_orb -> detectandcompute(d_srcr, mat(), keypoints_2, d_descriptorsr);
cout << "detect " << keypoints_1.size() << " and " << keypoints_2.size() << " keypoints " << endl;
// endtime = clock();
// cout << "took time : " << (double)(endtime - starttime) / clocks_per_sec * 1000 << " ms" << endl;
d_matcher -> match(d_descriptorsl, d_descriptorsr, matches);//l、r表示左右两幅图像进行匹配
//计算匹配所需时间
chrono::steady_clock::time_point t2 = chrono::steady_clock::now();
chrono::duration<double> time_used = chrono::duration_cast<chrono::duration<double>>(t2 - t1);
cout << "extract and match cost = " << time_used.count() * 1000 << " ms " << endl;
int sz = matches.size();
double max_dist = 0; double min_dist = 100;
for (int i = 0; i < sz; i++)
{
double dist = matches[i].distance;
if (dist < min_dist) min_dist = dist;
if (dist > max_dist) max_dist = dist;
}
for (int i = 0; i < sz; i++)
{
if (matches[i].distance < 0.6*max_dist)
{
good_matches.push_back(matches[i]);
}
}
cout << "match " << good_matches.size() << " keypoints " << endl;
// endtime = clock();
// cout << "took time : " << (double)(endtime - starttime) / clocks_per_sec * 1000 << " ms" << endl;
//提取良好匹配结果中在待测图片上的点集，确定匹配的大概位置
for (size_t i = 0; i < good_matches.size(); ++i)
{
scene.push_back(keypoints_2[ good_matches[i].trainidx ].pt);
}
for(unsigned int j = 0; j < scene.size(); j++)
cv::circle(img_2, scene[j], 2, cv::scalar(0, 255, 0), 2);
//画出普通匹配结果
mat showmatches;
drawmatches(img_1,keypoints_1,img_2,keypoints_2,matches,showmatches);
if (show_picture)
imshow("matches", showmatches);
// imwrite("matches.png", showmatches);
//画出良好匹配结果
mat showgoodmatches;
drawmatches(img_1,keypoints_1,img_2,keypoints_2,good_matches,showgoodmatches);
if (show_picture)
imshow("good_matches", showgoodmatches);
// imwrite("good_matches.png", showgoodmatches);
//画出良好匹配结果中在待测图片上的点集
if (show_picture)
imshow("matchpoints_in_img_2", img_2);
// imwrite("matchpoints_in_img_2.png", img_2);
if (show_picture)
waitkey(0);
}
int main(int argc, char **argv)
{
string picture1=string(argv[1]);
string picture2=string(argv[2]);
// string picture1 = "data/picture1/6.jpg";
// string picture2 = "data/picture2/16.png";
cout << "\nextract_orb::" << endl;
extract_orb(picture1, picture2);
cout << "\nextract_orb::" << endl;
extract_orb2(picture1, picture2);
cout << "\nextract_surf::" << endl;
extract_surf(picture1, picture2);
cout << "\nextract_akaze::" << endl;
extract_akaze(picture1, picture2);
cout << "\nextract_sift::" << endl;
extract_sift(picture1, picture2);
cout << "success!!" << endl;
}
cmakelists.txt
cmake_minimum_required(version 2.8.3) # 设定版本
project(descriptorcompare) # 设定工程名
set(cmake_cxx_compiler "g++") # 设定编译器
add_compile_options(-std=c++14) #编译选项，选择c++版本
# 设定可执行二进制文件的目录（最后生成的可执行文件放置的目录）
set(executable_output_path ${project_source_dir})
set(cmake_cxx_flags "${cmake_cxx_flags} -wall -fpermissive -g -o3 -wno-unused-function -wno-return-type")
find_package(opencv 3.0 required)
message(status "using opencv version ${opencv_version}")
find_package(eigen3 3.3.8 required)
find_package(pangolin required)
# 设定链接目录
link_directories(${project_source_dir}/lib)
# 设定头文件目录
include_directories(
${project_source_dir}/include
${eigen3_include_dir}
${opencv_include_dir}
${pangolin_include_dirs}
)
add_library(${project_name}
test.cc
)
target_link_libraries( ${project_name}
${opencv_libs}
${eigen3_libs}
${pangolin_libraries}
)
add_executable(main main.cpp )
target_link_libraries(main ${project_name} )
add_executable(icp icp.cpp )
target_link_libraries(icp ${project_name} )
执行效果
./main 1.png 2.png
extract_orb:: detect 1500 and 1500 keypoints extract and match cost = 21.5506 ms match 903 keypoints extract_orb:: detect 1304 and 1301 keypoints extract and match orb cost = 25.4976 ms match 313 keypoints extract_surf:: detect 915 and 940 keypoints extract and match cost = 53.8371 ms match 255 keypoints extract_sift:: detect 1536 and 1433 keypoints extract and match cost = 97.9322 ms match 213 keypoints success!!
icp
#include <iostream>
#include <opencv2/core/core.hpp>
#include <opencv2/features2d/features2d.hpp>
#include <opencv2/highgui/highgui.hpp>
#include <opencv2/calib3d/calib3d.hpp>
#include <eigen/core>
#include <eigen/dense>
#include <eigen/geometry>
#include <eigen/svd>
#include <pangolin/pangolin.h>
#include <chrono>
using namespace std;
using namespace cv;
int picture_h=480;
int picture_w=640;
bool show_picture = true;
void find_feature_matches(
const mat &img_1, const mat &img_2,
std::vector<keypoint> &keypoints_1,
std::vector<keypoint> &keypoints_2,
std::vector<dmatch> &matches);
// 像素坐标转相机归一化坐标
point2d pixel2cam(const point2d &p, const mat &k);
void pose_estimation_3d3d(
const vector<point3f> &pts1,
const vector<point3f> &pts2,
mat &r, mat &t
);
int main(int argc, char **argv) {
if (argc != 5) {
cout << "usage: pose_estimation_3d3d img1 img2 depth1 depth2" << endl;
return 1;
}
//-- 读取图像
mat img_1 = imread(argv[1], cv_load_image_color);
mat img_2 = imread(argv[2], cv_load_image_color);
vector<keypoint> keypoints_1, keypoints_2;
vector<dmatch> matches;
find_feature_matches(img_1, img_2, keypoints_1, keypoints_2, matches);
cout << "picture1 keypoints: " << keypoints_1.size() << " \npicture2 keypoints: " << keypoints_2.size() << endl;
cout << "一共找到了 " << matches.size() << " 组匹配点" << endl;
// 建立3d点
mat depth1 = imread(argv[3], cv_8uc1); // 深度图为16位无符号数，单通道图像
mat depth2 = imread(argv[4], cv_8uc1); // 深度图为16位无符号数，单通道图像
mat k = (mat_<double>(3, 3) << 595.2, 0, 328.9, 0, 599.0, 253.9, 0, 0, 1);
vector<point3f> pts1, pts2;
for (dmatch m:matches) {
int d1 = 255-(int)depth1.ptr<uchar>(int(keypoints_1[m.queryidx].pt.y))[int(keypoints_1[m.queryidx].pt.x)];
int d2 = 255-(int)depth2.ptr<uchar>(int(keypoints_2[m.trainidx].pt.y))[int(keypoints_2[m.trainidx].pt.x)];
if (d1 == 0 || d2 == 0) // bad depth
continue;
point2d p1 = pixel2cam(keypoints_1[m.queryidx].pt, k);
point2d p2 = pixel2cam(keypoints_2[m.trainidx].pt, k);
float dd1 = int(d1) / 1000.0;
float dd2 = int(d2) / 1000.0;
pts1.push_back(point3f(p1.x * dd1, p1.y * dd1, dd1));
pts2.push_back(point3f(p2.x * dd2, p2.y * dd2, dd2));
}
cout << "3d-3d pairs: " << pts1.size() << endl;
mat r, t;
pose_estimation_3d3d(pts1, pts2, r, t);
//dzq add
cv::mat pose = (mat_<double>(4, 4) << r.at<double>(0, 0), r.at<double>(0, 1), r.at<double>(0, 2), t.at<double>(0),
r.at<double>(1, 0), r.at<double>(1, 1), r.at<double>(1, 2), t.at<double>(1),
r.at<double>(2, 0), r.at<double>(2, 1), r.at<double>(2, 2), t.at<double>(2),
0, 0, 0, 1);
cout << "[delete outliers] matched objects distance: ";
vector<double> vdistance;
double alldistance = 0; //存储总距离，用来求平均匹配距离，用平均的误差距离来剔除外点
for (int i = 0; i < pts1.size(); i++)
{
mat point = pose * (mat_<double>(4, 1) << pts2[i].x, pts2[i].y, pts2[i].z, 1);
double distance = pow(pow(pts1[i].x - point.at<double>(0), 2) + pow(pts1[i].y - point.at<double>(1), 2) + pow(pts1[i].z - point.at<double>(2), 2), 0.5);
vdistance.push_back(distance);
alldistance += distance;
// cout << distance << " ";
}
// cout << endl;
double avgdistance = alldistance / pts1.size(); //求一个平均距离
int n_outliers = 0;
for (int i = 0, j = 0; i < pts1.size(); i++, j++) //i用来记录剔除后vector遍历的位置，j用来记录原位置
{
if (vdistance[i] > 1.5 * avgdistance) //匹配物体超过平均距离的n倍就会被剔除 [delete outliers] dzq fixed_param
{
n_outliers++;
}
}
cout << "n_outliers:: " << n_outliers << endl;
// show points
{
//创建一个窗口
pangolin::createwindowandbind("show points", 640, 480);
//启动深度测试
glenable(gl_depth_test);
// define projection and initial modelview matrix
pangolin::openglrenderstate s_cam(
pangolin::projectionmatrix(640, 480, 420, 420, 320, 240, 0.05, 500),
//对应的是glulookat,摄像机位置,参考点位置,up vector(上向量)
pangolin::modelviewlookat(0, -5, 0.1, 0, 0, 0, pangolin::axisy));
// create interactive view in window
pangolin::handler3d handler(s_cam);
//setbounds 跟opengl的viewport 有关
//看simpledisplay中边界的设置就知道
pangolin::view &d_cam = pangolin::createdisplay()
.setbounds(0.0, 1.0, 0.0, 1.0, -640.0f / 480.0f)
.sethandler(&handler);
while (!pangolin::shouldquit())
{
// clear screen and activate view to render into
glclearcolor(0.97,0.97,1.0, 1); //背景色
glclear(gl_color_buffer_bit | gl_depth_buffer_bit);
d_cam.activate(s_cam);
glbegin(gl_points); //绘制匹配点
gllinewidth(5);
for (int i = 0; i < pts1.size(); i++)
{
glcolor3f(1, 0, 0);
glvertex3d(pts1[i].x,pts1[i].y,pts1[i].z);
mat point = pose * (mat_<double>(4, 1) << pts2[i].x, pts2[i].y, pts2[i].z, 1);
glcolor3f(0, 1, 0);
glvertex3d(point.at<double>(0),point.at<double>(1),point.at<double>(2));
}
glend();
glbegin(gl_lines); //绘制匹配线
gllinewidth(1);
for (int i = 0; i < pts1.size(); i++)
{
glcolor3f(0, 0, 1);
glvertex3d(pts1[i].x,pts1[i].y,pts1[i].z);
mat point = pose * (mat_<double>(4, 1) << pts2[i].x, pts2[i].y, pts2[i].z, 1);
glvertex3d(point.at<double>(0),point.at<double>(1),point.at<double>(2));
}
glend();
glbegin(gl_points); //绘制所有点
gllinewidth(5);
glcolor3f(1, 0.5, 0);
for (int i = 0; i < picture_h; i+=2)
{
for (int j = 0; j < picture_w; j+=2)
{
int d1 = 255-(int)depth1.ptr<uchar>(i)[j];
if (d1 == 0) // bad depth
continue;
point2d temp_p;
temp_p.y=i; //这里的x和y应该和i j相反
temp_p.x=j;
point2d p1 = pixel2cam(temp_p, k);
float dd1 = int(d1) / 1000.0;
glvertex3d(p1.x * dd1, p1.y * dd1, dd1);
// glvertex3d(j/1000.0, i/1000.0, d1/200.0);
}
}
glend();
// swap frames and process events
pangolin::finishframe();
}
}
}
void find_feature_matches(const mat &img_1, const mat &img_2,
std::vector<keypoint> &keypoints_1,
std::vector<keypoint> &keypoints_2,
std::vector<dmatch> &matches) {
//-- 初始化
mat descriptors_1, descriptors_2;
// used in opencv3
ptr<featuredetector> detector = orb::create(2000,(1.200000048f), 8, 100);
ptr<descriptorextractor> descriptor = orb::create(5000);
ptr<descriptormatcher> matcher = descriptormatcher::create("bruteforce-hamming");
//-- 第一步:检测 oriented fast 角点位置
detector->detect(img_1, keypoints_1);
detector->detect(img_2, keypoints_2);
//-- 第二步:根据角点位置计算 brief 描述子
descriptor->compute(img_1, keypoints_1, descriptors_1);
descriptor->compute(img_2, keypoints_2, descriptors_2);
//-- 第三步:对两幅图像中的brief描述子进行匹配，使用 hamming 距离
vector<dmatch> match;
// bfmatcher matcher ( norm_hamming );
matcher->match(descriptors_1, descriptors_2, match);
//-- 第四步:匹配点对筛选
double min_dist = 10000, max_dist = 0;
//找出所有匹配之间的最小距离和最大距离, 即是最相似的和最不相似的两组点之间的距离
for (int i = 0; i < descriptors_1.rows; i++) {
double dist = match[i].distance;
if (dist < min_dist) min_dist = dist;
if (dist > max_dist) max_dist = dist;
}
printf("-- max dist : %f \n", max_dist);
printf("-- min dist : %f \n", min_dist);
//当描述子之间的距离大于两倍的最小距离时,即认为匹配有误.但有时候最小距离会非常小,设置一个经验值30作为下限.
for (int i = 0; i < descriptors_1.rows; i++) {
if (match[i].distance <= max(2 * min_dist, 30.0)) {
matches.push_back(match[i]);
}
}
//-- 第五步:绘制匹配结果
if(show_picture)
{
mat img_match;
mat img_goodmatch;
drawmatches(img_1, keypoints_1, img_2, keypoints_2, matches, img_match);
imshow("all matches", img_match);
waitkey(0);
}
}
point2d pixel2cam(const point2d &p, const mat &k) {
return point2d(
(p.x - k.at<double>(0, 2)) / k.at<double>(0, 0),
(p.y - k.at<double>(1, 2)) / k.at<double>(1, 1)
);
}
void pose_estimation_3d3d(const vector<point3f> &pts1,
const vector<point3f> &pts2,
mat &r, mat &t) {
point3f p1, p2; // center of mass
int n = pts1.size();
for (int i = 0; i < n; i++) {
p1 += pts1[i];
p2 += pts2[i];
}
p1 = point3f(vec3f(p1) / n);
p2 = point3f(vec3f(p2) / n);
vector<point3f> q1(n), q2(n); // remove the center
for (int i = 0; i < n; i++) {
q1[i] = pts1[i] - p1;
q2[i] = pts2[i] - p2;
}
// compute q1*q2^t
eigen::matrix3d w = eigen::matrix3d::zero();
for (int i = 0; i < n; i++) {
w += eigen::vector3d(q1[i].x, q1[i].y, q1[i].z) * eigen::vector3d(q2[i].x, q2[i].y, q2[i].z).transpose();
}
// cout << "w=" << w << endl;
// svd on w
eigen::jacobisvd<eigen::matrix3d> svd(w, eigen::computefullu | eigen::computefullv);
eigen::matrix3d u = svd.matrixu();
eigen::matrix3d v = svd.matrixv();
eigen::matrix3d r_ = u * (v.transpose());
if (r_.determinant() < 0) {
r_ = -r_;
}
eigen::vector3d t_ = eigen::vector3d(p1.x, p1.y, p1.z) - r_ * eigen::vector3d(p2.x, p2.y, p2.z);
// convert to cv::mat
r = (mat_<double>(3, 3) <<
r_(0, 0), r_(0, 1), r_(0, 2),
r_(1, 0), r_(1, 1), r_(1, 2),
r_(2, 0), r_(2, 1), r_(2, 2)
);
t = (mat_<double>(3, 1) << t_(0, 0), t_(1, 0), t_(2, 0));
}
void convertrgb2gray(string picture)
{
double min;
double max;
mat depth_new_1 = imread(picture); // 深度图为16位无符号数，单通道图像
mat test=mat(20,256,cv_8uc3);
int s;
for (int i = 0; i < 20; i++) {
std::cout<<i<<" ";
vec3b* p = test.ptr<vec3b>(i);
for (s = 0; s < 32; s++) {
p[s][0] = 128 + 4 * s;
p[s][1] = 0;
p[s][2] = 0;
}
p[32][0] = 255;
p[32][1] = 0;
p[32][2] = 0;
for (s = 0; s < 63; s++) {
p[33+s][0] = 255;
p[33+s][1] = 4+4*s;
p[33+s][2] = 0;
}
p[96][0] = 254;
p[96][1] = 255;
p[96][2] = 2;
for (s = 0; s < 62; s++) {
p[97 + s][0] = 250 - 4 * s;
p[97 + s][1] = 255;
p[97 + s][2] = 6+4*s;
}
p[159][0] = 1;
p[159][1] = 255;
p[159][2] = 254;
for (s = 0; s < 64; s++) {
p[160 + s][0] = 0;
p[160 + s][1] = 252 - (s * 4);
p[160 + s][2] = 255;
}
for (s = 0; s < 32; s++) {
p[224 + s][0] = 0;
p[224 + s][1] = 0;
p[224 + s][2] = 252-4*s;
}
}
cout<<"depth_new_1 :: "<<depth_new_1.cols<<" "<<depth_new_1.rows<<" "<<endl;
mat img_g=mat(picture_h,picture_w,cv_8uc1);
for(int i=0;i<picture_h;i++)
{
vec3b *p = test.ptr<vec3b>(0);
vec3b *q = depth_new_1.ptr<vec3b>(i);
for (int j = 0; j < picture_w; j++)
{
for(int k=0;k<256;k++)
{
if ( (((int)p[k][0] - (int)q[j][0] < 4) && ((int)q[j][0] - (int)p[k][0] < 4))&&
(((int)p[k][1] - (int)q[j][1] < 4) && ((int)q[j][1] - (int)p[k][1] < 4))&&
(((int)p[k][2] - (int)q[j][2] < 4) && ((int)q[j][2] - (int)p[k][2] < 4)))
{
img_g.at<uchar>(i,j)=k;
}
}
}
}
imwrite("14_depth_3.png", img_g);
waitkey();
}
cmakelists.txt
和上面一样。
./icp 1.png 2.png 1_depth.png 2_depth.png
-- max dist : 87.000000 -- min dist : 4.000000 picture1 keypoints: 1304 picture2 keypoints: 1301 一共找到了 313 组匹配点 3d-3d pairs: 313 [delete outliers] matched objects distance: n_outliers:: 23
执行效果
以上就是浅析orb、surf、sift特征点提取方法以及icp匹配方法的详细内容，更多关于特征点提取方法 icp匹配方法的资料请关注其它相关文章！