ORB算法特征匹配

ORB算法是brief算法的改进版。ORB算法比SIFT算法效率高两个数量级。计算速度上ORB是SIFT的100倍，是SURF的10倍。综合性能是最好的。
ORB算法没有解决brief算法的不具备尺度不变性的缺点。但是因为是追求速度的算法，常常适用于视屏流的处理，可以通过跟踪或者一些启发式的策略来解决尺度不变性的问题。
在给出的程序代码中，使用了基于FLANN的描述符对象匹配。
高维数据的快速最近邻算法FLANN
程序中使用的是LinearIndexParams，该结构对应的索引进行线性的、暴力(brute-force)的搜索。
其他的还有基于kd树的索引。
KNN算法与Kd树
kd树原理比较复杂。。。。。。以后有时间再看。。。。。。。。

#include<opencv2/opencv.hpp>#include<opencv2/highgui/highgui.hpp>#include<opencv2/features2d/features2d.hpp>#include<opencv2/xfeatures2d/nonfree.hpp>#include<iostream>using namespace std;using namespace cv;int main(){    Mat srcImage = imread("1.jpg",1);    imshow("原始图",srcImage);    Mat srcImage_gray;    cvtColor(srcImage,srcImage_gray,CV_BGR2GRAY);    Ptr<ORB> Detector = ORB::create(400);    Ptr<ORB> Extrator = ORB::create();    vector<KeyPoint>keyPoints;    Mat descriptor;    Detector->detect(srcImage_gray,keyPoints);    Extrator->compute(srcImage_gray,keyPoints,descriptor);    flann::Index flannIndex(descriptor,flann::LshIndexParams(12,20,2),cvflann::FLANN_DIST_HAMMING);    VideoCapture cap(0);    cap.set(CV_CAP_PROP_FRAME_WIDTH,360);    cap.set(CV_CAP_PROP_FRAME_HEIGHT,900);    unsigned int frameCount = 0;    while (1)    {        double time0 = static_cast<double>(getTickCount());        Mat captureImage, captureImage_gray;        cap >> captureImage;        if (captureImage.empty())            continue;        cvtColor(captureImage,captureImage_gray,CV_BGR2GRAY);        vector<KeyPoint>captureKeyPoints;        Mat captureDescriptor;        Detector->detect(captureImage_gray,captureKeyPoints);        Extrator->compute(captureImage_gray,captureKeyPoints,captureDescriptor);        Mat matchIndex(captureDescriptor.rows, 2, CV_32SC1);        Mat matchDistance(captureDescriptor.rows,2,CV_32FC1);        flannIndex.knnSearch(captureDescriptor,matchIndex,matchDistance,2,flann::SearchParams());        vector<DMatch>goodMatches;        for (int i = 0; i < matchDistance.rows; i++)        {            if (matchDistance.at<float>(i, 0) < 0.6*matchDistance.at<float>(i, 1))            {                DMatch dmarches(i, matchIndex.at<int>(i, 0), matchDistance.at<float>(i, 0));                goodMatches.push_back(dmarches);            }        }        Mat resultImage;        drawMatches(captureImage,captureKeyPoints,srcImage,keyPoints,goodMatches,resultImage);        imshow("匹配窗口",resultImage);        cout << "帧率：" << getTickFrequency() / (getTickCount() - time0) << endl;        if (char(waitKey(1) == 27))break;    }![这里写图片描述](http://img.blog.csdn.net/20170609181804860?watermark/2/text/aHR0cDovL2Jsb2cuY3Nkbi5uZXQveHh6eHh6ZGx1dA==/font/5a6L5L2T/fontsize/400/fill/I0JBQkFCMA==/dissolve/70/gravity/SouthEast)    return 0;}

可见帧率比起SURF算法要高出很多。
opencv3.2 SURF实现特征点匹配