【VS2015&Opencv3.4.1基于图像特征描述子的图像配准学习备忘笔记一】

   使用opencv对图像进行特征提取工作中经常用到surf与sift等相关特征描述子，该模块在opencv_contrib模块中，使用时候需要注意与对应的opencv版本进行编译。opencv2.x版本与opencv3.x中关于特征子的使用不太一样，这里根据最新的opencv版本来统一进行代码的设计执行。

  opencv3.2中SurfFeatureDetector、SurfDescriptorExtractor、BruteForceMatcher这三个的使用方法已经和原先2.4版本前不一样了。使用方法示例如下：

 Ptr<SURF> detector = SURF::create(minHessian);
 detector->detect(img_1, keypoints_1);

 Ptr<SURF> extractor = SURF::create();
 extractor->compute(img_1, keypoints_1, descriptors_1);

Ptr<DescriptorMatcher> matcher = DescriptorMatcher::create("BruteForce");//这里填写使用的匹配方式
matcher->match(descriptors_1, descriptors_2, matches);

   图像配准(Image registration)就是将不同时间、不同传感器（成像设备）或不同条件下（天候、照度、摄像位置和角度等）获取的两幅或多幅图像进行匹配、叠加的过程，它已经被广泛地应用于遥感数据分析、计算机视觉、图像处理等领域。

---

图像配准基本过程：

1、先拍摄两张有相同区域的图片，注意图片尺寸保持一致。

2、分别提取出图像的特征点(如果图像质量很差的话，可能需要先做些预处理操作)。

3、根据图像特征点，对它们做特征点匹配。

4、筛选出比较好的特征匹配点。

5、根据这些特征匹配点计算出畸变仿射矩阵。

6、使用算出来的矩阵进行图像匹配。

---

以下是opencv2.x的代码（参考链接https://blog.csdn.net/dcrmg/article/details/52627726）：

#include "highgui/highgui.hpp"  

#include "opencv2/nonfree/nonfree.hpp"  

#include "opencv2/legacy/legacy.hpp" 

#include <iostream>

 

using namespace cv;

using namespace std;

 

int main(int argc,char *argv[])  

{  

	Mat image01=imread(argv[1]);  

	Mat image02=imread(argv[2]);

	imshow("原始测试图像",image01);

	imshow("基准图像",image02);

 

	//灰度图转换

	Mat image1,image2;  

	cvtColor(image01,image1,CV_RGB2GRAY);

	cvtColor(image02,image2,CV_RGB2GRAY);

 

 

	//提取特征点  

	SurfFeatureDetector surfDetector(800);  // 海塞矩阵阈值

	vector<KeyPoint> keyPoint1,keyPoint2;  

	surfDetector.detect(image1,keyPoint1);  

	surfDetector.detect(image2,keyPoint2);	

 

	//特征点描述，为下边的特征点匹配做准备  

	SurfDescriptorExtractor SurfDescriptor;  

	Mat imageDesc1,imageDesc2;  

	SurfDescriptor.compute(image1,keyPoint1,imageDesc1);  

	SurfDescriptor.compute(image2,keyPoint2,imageDesc2);	

 

	//获得匹配特征点，并提取最优配对  	

	FlannBasedMatcher matcher;

	vector<DMatch> matchePoints;  

	matcher.match(imageDesc1,imageDesc2,matchePoints,Mat());

	sort(matchePoints.begin(),matchePoints.end()); //特征点排序	

 

	//获取排在前N个的最优匹配特征点

	vector<Point2f> imagePoints1,imagePoints2;	

	for(int i=0;i<10;i++)

	{		

		imagePoints1.push_back(keyPoint1[matchePoints[i].queryIdx].pt);		

		imagePoints2.push_back(keyPoint2[matchePoints[i].trainIdx].pt);		

	}

 

	//获取图像1到图像2的投影映射矩阵 尺寸为3*3

	Mat homo=findHomography(imagePoints1,imagePoints2,CV_RANSAC);

	////也可以使用getPerspectiveTransform方法获得透视变换矩阵，不过要求只能有4个点，效果稍差

	//Mat	homo=getPerspectiveTransform(imagePoints1,imagePoints2);   

        cout<<"变换矩阵为：\n"<<homo<<endl<<endl; //输出映射矩阵

	//图像配准

	Mat imageTransform1,imageTransform2;

	warpPerspective(image01,imageTransform1,homo,Size(image02.cols,image02.rows));	

	imshow("经过透视矩阵变换后",imageTransform1);

	

	waitKey();  

	return 0;  

}

本博客使用的opencv3.x版本代码如下：

#include<opencv2/opencv.hpp>

#include<opencv2/xfeatures2d/nonfree.hpp>
//在使用SurfFeatureDetector类时候，opencv3.x新版本需要加相关头文件与命名空间

#include <iostream>



using namespace cv;

using namespace std;

using namespace xfeatures2d;

int main(int argc, char *argv[])

{

	Ptr<SurfFeatureDetector> detector = SurfFeatureDetector::create(800);
	Mat image01 = imread("1.png");

	Mat image02 = imread("2.png");

	imshow("原始测试图像", image01);

	imshow("基准图像", image02);



	//灰度图转换

	Mat srcImage1, srcImage2;

	cvtColor(image01, srcImage1, CV_RGB2GRAY);

	cvtColor(image02, srcImage2, CV_RGB2GRAY);
	vector<cv::KeyPoint> key_points_1, key_points_2;

	Mat dstImage1, dstImage2;
	detector->detectAndCompute(srcImage1, Mat(), key_points_1, dstImage1);
	detector->detectAndCompute(srcImage2, Mat(), key_points_2, dstImage2);//可以分成detect和compute

	Mat img_keypoints_1, img_keypoints_2;
	drawKeypoints(srcImage1, key_points_1, img_keypoints_1, Scalar::all(-1), DrawMatchesFlags::DEFAULT);
	drawKeypoints(srcImage2, key_points_2, img_keypoints_2, Scalar::all(-1), DrawMatchesFlags::DEFAULT);

	Ptr<DescriptorMatcher> matcher = DescriptorMatcher::create("FlannBased");
	vector<DMatch>mach;

	matcher->match(dstImage1, dstImage2, mach);

	sort(mach.begin(), mach.end()); //特征点排序	
	double Max_dist = 0;
	double Min_dist = 100;
	for (int i = 0; i < dstImage1.rows; i++)
	{
		double dist = mach[i].distance;
		if (dist < Min_dist)Min_dist = dist;
		if (dist > Max_dist)Max_dist = dist;
	}
	cout << "最短距离" << Min_dist << endl;
	cout << "最长距离" << Max_dist << endl;

	vector<DMatch>goodmaches;
	for (int i = 0; i < dstImage1.rows; i++)
	{
		if (mach[i].distance < 2 * Min_dist)
			goodmaches.push_back(mach[i]);
	}
	Mat img_maches;
	drawMatches(srcImage1, key_points_1, srcImage2, key_points_2, goodmaches, img_maches);

	vector<Point2f> imagePoints1, imagePoints2;

	for (int i = 0; i<10; i++)

	{

		imagePoints1.push_back(key_points_1[mach[i].queryIdx].pt);

		imagePoints2.push_back(key_points_2[mach[i].trainIdx].pt);

	}



	Mat homo = findHomography(imagePoints1, imagePoints2, CV_RANSAC);

	////也可以使用getPerspectiveTransform方法获得透视变换矩阵，不过要求只能有4个点，效果稍差

	//Mat	homo=getPerspectiveTransform(imagePoints1,imagePoints2);   

	cout << "变换矩阵为：\n" << homo << endl << endl; //输出映射矩阵

												//图像配准

	Mat imageTransform1, imageTransform2;

	warpPerspective(image01, imageTransform1, homo, Size(image02.cols, image02.rows));

	imshow("经过透视矩阵变换后", imageTransform1);



	waitKey();

	return 0;

}

参考博客https://blog.csdn.net/u011630458/article/details/50561188

              https://blog.csdn.net/dcrmg/article/details/52627726

              https://blog.csdn.net/mangobar/article/details/80414695

              https://blog.csdn.net/xxzxxzdlut/article/details/72926011