OpenCV + CPP 系列（卌一）图像特征匹配（ FLANN 匹配）

文章目录

一、FLANN简介

FLANN库全称是Fast Library for Approximate Nearest Neighbors（快速最近邻逼进搜索库），是一个在高维空间快速搜索近邻的库，它是目前最完整的（近似）最近邻开源库。不但实现了一系列查找算法，还包含了一种自动选取最快算法的机制。

主要优势：实现了包含kd-tree等在内的搜索算法集合，并且可以根据数据集本身特点，自动选取最适合的算法来完成k近邻搜索任务（更加确切的说是高维特征向量的匹配任务），提供了c++、c、python、matlab的接口。

FLANN 算法 官网：​​https://www.cs.ubc.ca/paper/flann​​​ FLANN github地址：​​https:///flann-lib/flann​​

关于算法详情查看：​​javascript:void(0)​​​ 关于特征存储查看：​​OpenCV—python 角点特征检测之三（FLANN匹配）​​ 关于构造函数查看：​​https://www.jianshu.com/p/d70d9c8b2bec​​

主要介绍一下indexParams的参数

​​Ptr<flann::IndexParams> indexParams​​​; 索引参数
​​​Ptr<flann::SearchParams> searchParams​​​; 搜索参数
​​​Ptr<flann::Index> flannIndex​​; 创建索引

头文件 ​​image_feature_all.h​​：声明类与公共函数

#pragma once
#include <opencv2/opencv.hpp>
#include <iostream>
#include <opencv2/xfeatures2d.hpp>　　//新增引入库

using namespace cv;
using namespace std;


class ImageFeature {
public:
  void flann_demo(Mat& image, Mat& image2);

};

主函数​​main.cpp​​调用该类的公共成员函数

#include "image_feature_all.h"



int main(int argc, char** argv) {
  const char* img_path = "D:\\Desktop\\match_dst.jpg";
  const char* img_path2 = "D:\\Desktop\\match_raw.jpg";
  Mat image = imread(img_path, IMREAD_GRAYSCALE);
  Mat image2 = imread(img_path2, IMREAD_GRAYSCALE);
  if (image.empty() || image2.empty()) {
    cout << "图像数据为空，读取文件失败！" << endl;
  }
  ImageFeature imgfeature;
  imgfeature.flann_demo(image, image2);

  waitKey(0);
  destroyAllWindows();
  return 0;
}

演示SURF–Flann

源文件 ​​feature_extract.cpp​​：实现类与公共函数

void ImageFeature::flann_demo(Mat& image, Mat& image2) {
  //SURF 特征提取
  int minHessian = 400;
  Ptr<xfeatures2d::SURF> detector = xfeatures2d::SURF::create(minHessian);

  vector<KeyPoint> keypoint_obj, keypoint_scene;
  Mat descriptor_obj, descriptor_scens;
  detector->detectAndCompute(image, Mat(), keypoint_obj, descriptor_obj);
  detector->detectAndCompute(image2, Mat(), keypoint_scene, descriptor_scens);

  //SURF 特征匹配
  FlannBasedMatcher matcher;
  vector<DMatch> matches;
  matcher.match(descriptor_obj, descriptor_scens, matches);

  //查找最大最小距离
  double minDist = 1000;
  double maxDist = 0;
  for (size_t i = 0; i < descriptor_obj.rows; i++) {
    double dist = matches[i].distance;
    if (dist > maxDist) { maxDist = dist; }
    if (dist < minDist) { minDist = dist; }
  }
  cout << "minDist = " << minDist << endl;
  cout << "maxDist = " << maxDist << endl;

  // 筛选最佳匹配（小于最小距离的3倍）
  vector<DMatch> goodMatchs;
  for (size_t i = 0; i < descriptor_obj.rows; i++) {
    double dist = matches[i].distance;
    if (dist < max(3 * minDist, 0.02)) {
      goodMatchs.push_back(matches[i]);
    }
  }

  //绘制特征点
  Mat matchImg;
  drawMatches(
    image, keypoint_obj,
    image2, keypoint_scene, 
    goodMatchs, matchImg,
    Scalar::all(-1), Scalar::all(-1), 
    vector<char>(), DrawMatchesFlags::NOT_DRAW_SINGLE_POINTS
  );
  imshow("matchImg", matchImg);

  //查看goodMatchs
  for (size_t i = 0; i < goodMatchs.size(); i++){
    cout << "goodMatchs "<<to_string(i) << "\tKeypoint1, Keypoint2 = " << goodMatchs[i].queryIdx <<", " << goodMatchs[i].trainIdx << endl;
  }
}

二、单应性矩阵

从场景图像上得到特征匹配图像的四点坐标：​​详情请查看​​

函数介绍：findHomography有三个重载函数，这里只说明一个

Mat

InputArray srcPoints, 　　　　　　　特征点集合，一般是来自目标图像
InputArray dstPoints,　　　　　　　 特征点集合，一般是来自场景图像
int method = 0, 　　　　　　　　　　配准方法，支持有四种方法（如下）
double ransacReprojThreshold = 3,　　　　　　
OutputArray mask=noArray(), 　　　　掩码
const int maxIters = 2000,　　　　　　最大迭代次数，当使用RANSAC方法
const double confidence = 0.995　　　置信参数
);

0 – 使用所有的点，比如最小二乘

RANSAC – 基于随机样本一致性

LMEDS – 最小中值

RHO –基于渐近样本一致性

---

InputArray src, 　　 　　输入四个顶点

OutputArray dst, 　　 　 输出四个顶点（变换后）

InputArray m 　　　　　仿射变换矩阵

)

void ImageFeature::flann_demo(Mat& image, Mat& image2) {
  //SURF 特征提取
  int minHessian = 400;
  Ptr<xfeatures2d::SURF> detector = xfeatures2d::SURF::create(minHessian);

  vector<KeyPoint> keypoint_obj, keypoint_scene;
  Mat descriptor_obj, descriptor_scens;
  detector->detectAndCompute(image, Mat(), keypoint_obj, descriptor_obj);
  detector->detectAndCompute(image2, Mat(), keypoint_scene, descriptor_scens);

  //SURF 特征匹配
  FlannBasedMatcher matcher;
  vector<DMatch> matches;
  matcher.match(descriptor_obj, descriptor_scens, matches);

  //查找最大最小距离
  double minDist = 1000;
  double maxDist = 0;
  for (size_t i = 0; i < descriptor_obj.rows; i++) {
    double dist = matches[i].distance;
    if (dist > maxDist) { maxDist = dist; }
    if (dist < minDist) { minDist = dist; }
  }
  cout << "minDist = " << minDist << endl;
  cout << "maxDist = " << maxDist << endl;

  // 筛选最佳匹配（小于最小距离的3倍）
  vector<DMatch> goodMatchs;
  for (size_t i = 0; i < descriptor_obj.rows; i++) {
    double dist = matches[i].distance;
    if (dist < max(3 * minDist, 0.02)) {
      goodMatchs.push_back(matches[i]);
    }
  }

  //绘制特征点
  Mat matchImg;
  drawMatches(
    image, keypoint_obj,
    image2, keypoint_scene, 
    goodMatchs, matchImg,
    Scalar::all(-1), Scalar::all(-1), 
    vector<char>(), DrawMatchesFlags::NOT_DRAW_SINGLE_POINTS
  );
  imshow("matchImg", matchImg);

  //查看goodMatchs
  for (size_t i = 0; i < goodMatchs.size(); i++){
    cout << "goodMatchs "<<to_string(i) << "\tKeypoint1, Keypoint2 = " << goodMatchs[i].queryIdx <<", " << goodMatchs[i].trainIdx << endl;
  }

  //获取goodMatchs坐标,并获取变换矩阵
  vector<Point2f> obj;
  vector<Point2f> scens;
  for (size_t i = 0; i < goodMatchs.size(); i++) {
    obj.push_back(keypoint_obj[goodMatchs[i].queryIdx].pt);
    scens.push_back(keypoint_scene[goodMatchs[i].trainIdx].pt);
  }
  Mat H = findHomography(obj, scens, RANSAC);

  vector<Point2f> obj_corners(4);
  vector<Point2f> scens_corners(4);
  obj_corners[0] = Point(0, 0);
  obj_corners[1] = Point(image.cols, 0);
  obj_corners[2] = Point(image.cols,image.rows);
  obj_corners[3] = Point(0, image.rows);
  perspectiveTransform(obj_corners, scens_corners, H);

  //可视化
  line(matchImg, scens_corners[0] + obj_corners[1], scens_corners[1] + obj_corners[1], Scalar(0, 0, 255));
  line(matchImg, scens_corners[1] + obj_corners[1], scens_corners[2] + obj_corners[1], Scalar(0, 0, 255));
  line(matchImg, scens_corners[2] + obj_corners[1], scens_corners[3] + obj_corners[1], Scalar(0, 0, 255));
  line(matchImg, scens_corners[3] + obj_corners[1], scens_corners[0] + obj_corners[1], Scalar(0, 0, 255));
  imshow("matchImg2", matchImg);

  //绘制到背景图上
  Mat image2_BGR;
  cvtColor(image2, image2_BGR, COLOR_GRAY2BGR);
  line(image2_BGR, scens_corners[0], scens_corners[1], Scalar(0, 0, 255));
  line(image2_BGR, scens_corners[1], scens_corners[2], Scalar(0, 0, 255));
  line(image2_BGR, scens_corners[2], scens_corners[3], Scalar(0, 0, 255));
  line(image2_BGR, scens_corners[3], scens_corners[0], Scalar(0, 0, 255));
  imshow("image2_BGR", image2_BGR);
}