opencv3中SIFT配合暴力匹配进行关键点描述和提取

#include<opencv2/opencv.hpp>
#include<iostream>
#include<vector>

using namespace cv;
using namespace std;

int main()
{
Mat srcImage = imread("mofang1.jpg");
imshow("【原图】", srcImage);

//对BGR空间的图像直接进行计算很费时间，所以，需要转换为灰度图
Mat srcGrayImage;
cvtColor(srcImage, srcGrayImage, CV_BGR2GRAY);

//首先对两幅图像进行特征点的检测和描述子的计算
vector<KeyPoint> keyPoint1;
//这里用SURF会更加快
SIFT surf(2000);
surf.detect(srcGrayImage, keyPoint1);
Mat descriImage1;
surf.compute(srcGrayImage, keyPoint1, descriImage1);

//先对原图的描述子进行保留
BFMatcher FLMatcher;
//因为FlannBasedMatcher类的成员函数add()的参数是一个vector<Mat>的容器，所以先定义一个这样的变量，并将原图的描述子放入容器中
vector<Mat> g_vdescriImage1(1, descriImage1);
/*g_vdescriImage1.push_back(descriImage1);*/
//调用FlannBasedMatcher类的成员函数add，将原图的描述子放在FlannBasedMatcher的对象FLMatcher中
FLMatcher.add(g_vdescriImage1);
//...........................................................
FLMatcher.train();

VideoCapture capture;
capture.open(0);

Mat frameImage, frameGrayImage;
while (waitKey(1) != 27)
{
capture >> frameImage;

//为了提高计算效率，将图像转换为灰度图像
cvtColor(frameImage, frameGrayImage, CV_BGR2GRAY);

//计算特征点和描述子
vector<KeyPoint> keyPoints2;
surf.detect(frameGrayImage, keyPoints2);
Mat descriImage2;
surf.compute(frameGrayImage, keyPoints2, descriImage2);

//将之前得到的原图的描述子和现在得到的描述子进行匹配
//成员函数knnMatch的参数是二维的DMatch向量，所以首先定义一个该容器的向量
vector<vector<DMatch>> knnDMatches;
//用之前已经存放原图描述子的对象来计算
FLMatcher.knnMatch(descriImage2, knnDMatches, 2);

//采集优秀的匹配点
vector<DMatch> goodMatches;
for (size_t i = 0; i < knnDMatches.size(); i++)
{
//........................................................................
if (knnDMatches[i][0].distance < 0.6 * knnDMatches[i][1].distance)
{
goodMatches.push_back(knnDMatches[i][0]);
}
}

Mat dstImage;
drawMatches(frameImage, keyPoints2, srcImage, keyPoint1, goodMatches, dstImage);

imshow("【结果图】", dstImage);
}

return 0;
}