OpenCV4实现ORB特征匹配

工程配置

采用CLion+Ubuntu20.04进行开发，新建工程Study。

工程结构

结构如下：

.
├── CMakeLists.txt
├── include
│   └── main.h
└── src
    ├── CMakeLists.txt
    └── main.cpp

子目录include用于存放头文件；子目录src用于存放源码

CMake配置

根目录下CMakeLists.txt文件内容如下：

# cmake version
cmake_minimum_required(VERSION 3.21)
# project name
project(Study)
# cpp version
set(CMAKE_CXX_STANDARD 14)
# opencv
find_package(OpenCV REQUIRED)
include_directories(${OpenCV_INCLUDE_DIRS})
# incldue
include_directories(include)
# src
add_subdirectory(src)

src目录下CMakeLists.txt文件内容如下：

# exec
add_executable(Study main.cpp)
# link opencv
target_link_libraries(Study ${OpenCV_LIBS})

头文件配置

include目录中头文件main.h内容如下：

#ifndef STUDY_MAIN_H
#define STUDY_MAIN_H

#include 
#include 
#include 
//  OpenCV
#include 
#include 
#include 
#include 

//  namespace
using namespace std;
using namespace cv;

#endif //STUDY_MAIN_H

源文件

src下源文件main.cpp初始内容如下：

#include "main.h"

int main()
{

    return 0;
}

工程主代码

#include "main.h"

int main(){

    /*      读取图片    */
    Mat img_1 = imread("/home/jasonli/workspace/slambook2/ch7/1.png");
    Mat img_2 = imread("/home/jasonli/workspace/slambook2/ch7/2.png");
    assert(img_1.data != nullptr && img_2.data != nullptr);

    /*      初始化    */
    //  oriented fast 关键点数据
    std::vector<KeyPoint> keyPoints_1, keyPoints_2;
    //  brief 描述子数据
    Mat descriptors_1, descriptors_2;
    //  用于指向寻找关键点的类
    Ptr<FeatureDetector> detector = ORB::create();
    //  用于指向计算描述子的类
    Ptr<DescriptorExtractor> descriptor = ORB::create();
    //  用于指向特征匹配的类
    Ptr<DescriptorMatcher> matcher = DescriptorMatcher::create("BruteForce-Hamming");

    /*     ORB特征获取    */
    chrono::steady_clock::time_point t1 = chrono::steady_clock::now();
    //  寻找关键点
    detector->detect(img_1, keyPoints_1);
    detector->detect(img_2, keyPoints_2);
    //  计算描述子
    descriptor->compute(img_1, keyPoints_1, descriptors_1);
    descriptor->compute(img_2, keyPoints_2, descriptors_2);
    //  获取ORB特征消耗时间
    chrono::steady_clock::time_point  t2 = chrono::steady_clock::now();
    chrono::duration<double> time_used = chrono::duration_cast<chrono::duration<double>>(t2 - t1);
    cout << "extract ORB cost = " << time_used.count() << " seconds. " << endl;
    //  查看图像关键点信息
    Mat outImg1;
    drawKeypoints(img_1, keyPoints_1, outImg1);
    imshow("ORB Features", outImg1);
    imwrite("/home/jasonli/pic/ORB_Features.png", outImg1);
    /*     ORB特征匹配    */
    //  存储数据
    vector<DMatch> matches;
    t1 = chrono::steady_clock::now();
    //  对BRIEF描述子进行匹配
    matcher->match(descriptors_1, descriptors_2, matches);
    //  获取特征匹配的消耗时间
    t2 = chrono::steady_clock::now();
    time_used = chrono::duration_cast<chrono::duration<double>>(t2 - t1);
    cout << "match ORB cost = " << time_used.count() << " seconds. " << endl;

    /*     匹配点对筛选    */
    //  计算最大最小值，汉明距离
    auto min_max = minmax_element(matches.begin(), matches.end(),
                                  [](const DMatch &m1, const DMatch &m2){return m1.distance < m2.distance; });
    double min_dist = min_max.first->distance;
    double max_dist = min_max.second->distance;

    cout << "Max dist:\t" << max_dist << "\nMin dist:\t" << min_dist << endl;

    //  认定描述子间距大于最小距离两倍时，匹配错误
    vector<DMatch> good_matches;
    for(int i=0; i < descriptors_1.rows; i++){
        //  为防止最小间距过小，设置经验数据30
        if(matches[i].distance <= max(2 * min_dist , 30.0)){
            good_matches.push_back(matches[i]);
        }
    }

    /*     绘制匹配结果    */
    //  未筛选匹配结果
    Mat img_match;
    drawMatches(img_1, keyPoints_1, img_2, keyPoints_2, matches, img_match);
    imshow("all matches", img_match);
    imwrite("/home/jasonli/pic/all_matches.png", img_match);

    //  筛选后匹配结果
    Mat img_goodMatch;
    drawMatches(img_1, keyPoints_1, img_2, keyPoints_2, good_matches, img_goodMatch);
    imshow("good matches", img_goodMatch);
    imwrite("/home/jasonli/pic/good_matches.png", img_goodMatch);

    waitKey(0);
    return 0;
}

输出

内容如下：

extract ORB cost = 0.053812 seconds. 
match ORB cost = 0.000748446 seconds. 
Max dist:	94
Min dist:	4

特征点图像如下：

未筛选前的匹配如下：

筛选后的匹配如下：