Opencv (C++) learning series --- feature point detection and matching

The specific principles of feature detection and matching will be explained in detail in subsequent articles. This article mainly introduces the simple process of Opencv implementation:
Step 1: Define feature detectors (SIFT, SURF, ORB, etc.).

The second step: detect the feature points in the image, and store the feature points in Keypoints.

The third step: extract the description information of the feature points.

Step 4: Define a feature matcher (there are two main methods of feature matching, BFmatch and FlannBased).

Step 5: Filter out poor matching points (generally filter according to the distance between two adjacent points)

It is mainly filtered according to the distance in DMatch. Distance can be abstractly understood as a matching score. The smaller the distance, the higher the similarity of the detection points and the better the effect.

Step 6: Display the matched feature points.

Code 1 (unfiltered, only limited to 20 screening points)

#include <iostream>
#include <opencv2/opencv.hpp>  
#include<opencv2\highgui\highgui.hpp>
#include<opencv2\imgproc\imgproc.hpp>
#include <opencv2/xfeatures2d.hpp>


using namespace cv;  //包含cv命名空间
using namespace std;
using namespace xfeatures2d;


int main() {

	system("color 2E");
	//载入图片
	Mat src1 = imread("E:\\乔大花进度\\11-18\\sift特征检测和匹配\\3.jpg",1);
	Mat src2 = imread("E:\\乔大花进度\\11-18\\sift特征检测和匹配\\4.jpg", 1);

	//显示原图
	imshow("原图1",src1);
	imshow("原图2", src2);

	//定义变量
	vector<KeyPoint> keypoints1, keypoints2;//定义检测的特征点存储容器
	Mat descriptors1,descriptors2;//定义特征点描述信息为Mat类型
	Mat result_img;//匹配结果图片

	//创建sift特征检测器实例
	//将SIFT可以换位SURF、ORB
	Ptr<SIFT>detector = SIFT::create();
	//提取特征点
	detector->detect(src1,keypoints1,noArray());
	detector->detect(src2, keypoints2, Mat());


	//获取特征点的描述信息=>特征向量
	detector->compute(src1,keypoints1,descriptors1);
	detector->compute(src2, keypoints2, descriptors2);


	//定义匹配器的实例化=>方法为暴力匹配法
	Ptr<DescriptorMatcher> matcher = DescriptorMatcher::create(DescriptorMatcher::BRUTEFORCE);//create中的参数可以填string FlannBased等匹配方法

	//第二种实例化方法
	//BFMatcher matcher;

	//进行暴力匹配
	vector<DMatch> matches;

	//第一个参数为queryDescription为目标，第二个参数为trainDescription模板
	matcher->match(descriptors1,descriptors2,matches);


	//限制特征点匹配数量=》只匹配前20个较好的特征点
	int num = 20;
	nth_element(matches.begin(), matches.begin()+num,matches.end());
	//vector去除20以后的元素
	matches.erase(matches.begin()+num,matches.end());
	

	//输出关键点和匹配结果
	//其中右侧图为trainDescription模板，左侧图为queryDescription目标
	//左图中的点与右图中进行匹配对应
	drawMatches(src1,keypoints1,src2,keypoints2, matches,result_img);
	drawKeypoints(src1,keypoints1,src1);
	drawKeypoints(src2,keypoints2,src2);
	
	imshow("匹配结果",result_img);
	imshow("特征点1",src1);
	imshow("特征点2",src2);

	waitKey(0);
	system("pause");
	return 0;
}

The result of the operation is:

Code 2 (filtering by distance)

#include <iostream>
#include <opencv2/opencv.hpp>  
#include<opencv2\highgui\highgui.hpp>
#include<opencv2\imgproc\imgproc.hpp>
#include <opencv2/xfeatures2d.hpp>


using namespace cv;  //包含cv命名空间
using namespace std;
using namespace xfeatures2d;


int main() {

	system("color 2E");
	//载入图片
	Mat src1 = imread("E:\\乔大花进度\\11-18\\sift特征检测和匹配\\3.jpg",1);
	Mat src2 = imread("E:\\乔大花进度\\11-18\\sift特征检测和匹配\\4.jpg", 1);

	//显示原图
	imshow("原图1",src1);
	imshow("原图2", src2);

	//定义变量
	vector<KeyPoint> keypoints1, keypoints2;//定义检测的特征点存储容器
	Mat descriptors1,descriptors2;//定义特征点描述信息为Mat类型
	Mat result_img;//匹配结果图片

	//创建sift特征检测器实例
	//将SIFT可以换位SURF、ORB
	Ptr<SIFT>detector = SIFT::create();
	//提取特征点
	detector->detect(src1,keypoints1,noArray());
	detector->detect(src2, keypoints2, Mat());


	//获取特征点的描述信息=>特征向量
	detector->compute(src1,keypoints1,descriptors1);
	detector->compute(src2, keypoints2, descriptors2);


	//定义匹配器的实例化=>方法为暴力匹配法
	Ptr<DescriptorMatcher> matcher = DescriptorMatcher::create(DescriptorMatcher::BRUTEFORCE);//create中的参数可以填string FlannBased等匹配方法

	//第二种实例化方法
	//BFMatcher matcher;

	//进行暴力匹配
	vector<DMatch> matches;

	//第一个参数为queryDescription为目标，第二个参数为trainDescription模板
	matcher->match(descriptors1,descriptors2,matches);


	//限制特征点匹配数量=》只匹配前20个较好的特征点
	int num = 20;
	nth_element(matches.begin(), matches.begin()+num,matches.end());
	//vector去除20以后的元素
	matches.erase(matches.begin()+num,matches.end());
	
	double Max_distance = matches[1].distance;
	double Min_distance = matches[1].distance;
	vector<DMatch> goodfeatrues;
	
	//根据特征点的距离去筛选
	for (int i = 0; i < matches.size(); i++)
	{
		double dist = matches[i].distance;
		if (dist>Max_distance)
		{
			Max_distance = dist;
		}
		if (dist<Min_distance)
		{
			Min_distance = dist;
		}
		
	}

	cout << "匹配点的最大距离：" << Max_distance << endl;
	cout << "匹配点的最小距离：" << Min_distance << endl;

	//M为距离阈值，M越大点数越多
	double M = 1.3;
	for (int  i = 0; i < matches.size(); i++)
	{
		double dist = matches[i].distance;
		if (dist<M*Min_distance)		{
			goodfeatrues.push_back(matches[i]);
		}
	}
	cout << "最终选取特征点的数量为：" << matches.size() << endl;

	//输出关键点和匹配结果
	//其中右侧图为trainDescription模板，左侧图为queryDescription目标
	//左图中的点与右图中进行匹配对应
	drawMatches(src1,keypoints1,src2,keypoints2, goodfeatrues,result_img);
	drawKeypoints(src1,keypoints1,src1);
	drawKeypoints(src2,keypoints2,src2);
	
	imshow("匹配结果",result_img);
	imshow("特征点1",src1);
	imshow("特征点2",src2);

	waitKey(0);
	system("pause");
	return 0;
}

The result of the operation is:

 Code 3 (through knnMatch matching, you can filter by setting a threshold for distance, the effect is the best)

#include <iostream>
#include <opencv2/opencv.hpp>  
#include<opencv2\highgui\highgui.hpp>
#include<opencv2\imgproc\imgproc.hpp>
#include <opencv2/xfeatures2d.hpp>


using namespace cv;  //包含cv命名空间
using namespace std;
using namespace xfeatures2d;


int main() {

	system("color 2E");
	//载入图片
	Mat src1 = imread("E:\\乔大花进度\\11-18\\sift特征检测和匹配\\3.jpg",1);
	Mat src2 = imread("E:\\乔大花进度\\11-18\\sift特征检测和匹配\\4.jpg", 1);

	//显示原图
	imshow("原图1",src1);
	imshow("原图2", src2);

	//定义变量
	vector<KeyPoint> keypoints1, keypoints2;//定义检测的特征点存储容器
	Mat descriptors1,descriptors2;//定义特征点描述信息为Mat类型
	Mat result_img;//匹配结果图片

	//创建sift特征检测器实例
	//将SIFT可以换位SURF、ORB
	Ptr<SIFT>detector = SIFT::create();
	//提取特征点
	detector->detect(src1,keypoints1,noArray());
	detector->detect(src2, keypoints2, Mat());


	//获取特征点的描述信息=>特征向量
	detector->compute(src1,keypoints1,descriptors1);
	detector->compute(src2, keypoints2, descriptors2);


	//定义匹配器的实例化=>方法为暴力匹配法
	Ptr<DescriptorMatcher> matcher = DescriptorMatcher::create(DescriptorMatcher::BRUTEFORCE);//create中的参数可以填string FlannBased等匹配方法

	//第二种实例化方法
	//BFMatcher matcher;

	//进行暴力匹配
	vector<DMatch> matches;

	vector<Mat>train_desc(1, descriptors2);
	matcher->add(train_desc);
	matcher->train();

	vector<vector<DMatch>> matchpoints;
	matcher->knnMatch(descriptors1,matchpoints,2);

	vector<DMatch> goodfeatur;
	for (int i = 0; i < matchpoints.size(); i++)
	{
		if (matchpoints[i][0].distance<0.15*matchpoints[i][1].distance)
		{
			goodfeatur.push_back(matchpoints[i][0]);
		}

	}
	cout << "筛选后的特征点数量为： " << goodfeatur.size() << endl;

	//输出关键点和匹配结果
	//其中右侧图为trainDescription模板，左侧图为queryDescription目标
	//左图中的点与右图中进行匹配对应
	drawMatches(src1,keypoints1,src2,keypoints2, goodfeatur,result_img);
	drawKeypoints(src1,keypoints1,src1);
	drawKeypoints(src2,keypoints2,src2);
	
	namedWindow("匹配结果",WINDOW_NORMAL);
	resizeWindow("匹配结果",500,500);
	imshow("匹配结果",result_img);

	waitKey(0);
	system("pause");
	return 0;
}

The result of the operation is: