ORB对齐

//#include "pch.h"   //opencv4.0要去掉
#include <iostream>
#include <stdlib.h>
#include <opencv2/imgproc.hpp>
#include <opencv2/opencv.hpp>
#include "opencv2/xfeatures2d.hpp"
#include "opencv2/features2d.hpp"

using namespace std;
using namespace cv;
using namespace cv::xfeatures2d;

//最大特征点数
const int MAX_FEATURES = 500;
//好的特征点数
const float GOOD_MATCH_PERCENT = 0.15f;

/**
* @brief 图像对齐
*
* @param im1 对齐图像
* @param im2 模板图像
* @param im1Reg 输出图像
* @param h
*/
void alignImages(Mat &im1, Mat &im2, Mat &im1Reg, Mat &h) //引用传递，传入是实参的的别名，实际是实参的地址，因此调用被调函数后，实参会被间接修改。，这么看：Mat& arg
//和函数的 值传递 相反。
{
    // Convert images to grayscale
    Mat im1Gray, im2Gray;
    //转换为灰度图
    cvtColor(im1, im1Gray, COLOR_BGR2GRAY);   //opencv4.0这里CV_BGR2GRAY要改成COLOR_BGR2GRAY
    cvtColor(im2, im2Gray, COLOR_BGR2GRAY);

    // Variables to store keypoints and descriptors
    //关键点
    std::vector<KeyPoint> keypoints1, keypoints2;//定义装有KeyPoint数据类型的的数组变量， keypoints1, keypoints2。如：vector<int> x,y
    //特征描述符
    Mat descriptors1, descriptors2;

    // Detect ORB features and compute descriptors. 计算ORB特征和描述子
    Ptr<Feature2D> orb = ORB::create(MAX_FEATURES);//opencv中ORB函数是纯虚函数，继承自feature2D,无法通过实例化对象调用，只能通过指针调用。
    orb->detectAndCompute(im1Gray, Mat(), keypoints1, descriptors1);
    orb->detectAndCompute(im2Gray, Mat(), keypoints2, descriptors2);

    // Match features. 特征点匹配
    std::vector<DMatch> matches;
    //汉明距离进行特征点匹配
    Ptr<DescriptorMatcher> matcher = DescriptorMatcher::create("BruteForce-Hamming");//虚函数调用
    matcher->match(descriptors1, descriptors2, matches, Mat());

    // Sort matches by score 按照特征点匹配结果从优到差排列
    std::sort(matches.begin(), matches.end());

    // Remove not so good matches 移除不好的特征点
    const int numGoodMatches = matches.size() * GOOD_MATCH_PERCENT;
    matches.erase(matches.begin() + numGoodMatches, matches.end());//0.15*matchPoints_size

    // Draw top matches
    Mat imMatches;
    //画出特征点匹配图
    drawMatches(im1, keypoints1, im2, keypoints2, matches, imMatches);
    imwrite("matches.jpg", imMatches);             //暴力匹配后剔除差的匹配点后 打印的特征点匹配关系图

                                                   // Extract location of good matches
    std::vector<Point2f> points1, points2;

    //保存对应点
    for (size_t i = 0; i < matches.size(); i++)   //这里是暴力匹配作剔除后剩下的点
    {
        //queryIdx是对齐图像的描述子和特征点的下标。
        points1.push_back(keypoints1[matches[i].queryIdx].pt);
        //queryIdx是是样本图像的描述子和特征点的下标。
        points2.push_back(keypoints2[matches[i].trainIdx].pt);
    }

    // Find homography 计算Homography，RANSAC随机抽样一致性算法
    h = findHomography(points1, points2, RANSAC);//这个函数参数远比这个多，很多省略了，输入两张图片各自匹配点的矩阵，和匹配方法

    // Use homography to warp image 映射   
    warpPerspective(im1, im1Reg, h, im2.size()); //弯曲视角，（待配准图，结果图，单应矩阵，）
}

int main()
{
    // Read reference image 读取参考图像
    string refFilename("D://aa.jpg");        //基准图像
    cout << "Reading reference image : " << refFilename << endl;
    Mat imReference = imread(refFilename);

    // Read image to be aligned 读取对准图像
    string imFilename("D://a.jpg");   //待校准图像
    cout << "Reading image to align : " << imFilename << endl;
    Mat im = imread(imFilename);

    // Registered image will be resotred in imReg.
    // The estimated homography will be stored in h.
    //结果图像，单应性矩阵
    Mat imReg, h;

    // Align images
    cout << "Aligning images ..." << endl;
    alignImages(im, imReference, imReg, h);//（待配准图，参照图，结果图，单应矩阵）

    // Write aligned image to disk.
    string outFilename("aligned.jpg");   
    cout << "Saving aligned image : " << outFilename << endl;
    imwrite(outFilename, imReg);  //校准后图像

    /* Print estimated homography          
    [0.4260230318277862, 2.205599317460628, 7.436838081684622;
    -0.8418214464779903, 1.649310552258939, 1202.559886226323;
    0.0003634828675145606, 0.001245337973396314, 1]*/
    cout << "Estimated homography : \n" << h << endl; //这里输出到控制台了，要想保存需要输出重定向
    system("pause");
    return 0;
}