视觉里程计01 - ORB特征提取与跟踪

orb特征

概念网上描述的很详细，这里简单说一下

• 采用改进的FAST特征点，在FAST特征点提取的基础上加入多层金字塔来确定不同尺度下的特征点
• 用ID3方法来确定最优特征点
• 采用非极大值抑制去除局部较密集特征点
• 尺度不变性用金字塔来确定，旋转不变性用图像质心的夹角来确定
• 描述子采用BRIEF

总的来说该算法比surf还要快，而且准确率也很不错。

orb特征提取与跟踪代码

主体部分来自高翔的视觉SLAM14讲的代码，在开头加入了摄像头读取的代码替换掉固定的图片演示。这里用双目摄像头来进行匹配，实际上视觉里程计采用单目即可。

#include <opencv2/core.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/videoio.hpp>
#include <iostream>
#include "opencv2/features2d/features2d.hpp"
#include <vector>
#include <time.h>

using namespace cv;
using namespace std;

int main()
{
    VideoCapture cap1;
    VideoCapture cap2;
    cap1.open(1);//白色摄像头
    cap2.open(2);//黑色摄像头
    if (!cap1.isOpened()||!cap2.isOpened())
    {
        return -1;
    }
    //将摄像头从640*480改成320*240，速度从200ms提升至50ms
    cap1.set(CV_CAP_PROP_FRAME_WIDTH, 320);
    cap1.set(CV_CAP_PROP_FRAME_HEIGHT, 240);
    cap2.set(CV_CAP_PROP_FRAME_WIDTH, 320);
    cap2.set(CV_CAP_PROP_FRAME_HEIGHT, 240);
    //namedWindow("Video", 1);
    //namedWindow("Video", 2);
    //namedWindow("pts", 3);
    //Mat frame;
    
    Mat img_1;
    Mat img_2;
    while (1)
    {
        cap1 >> img_1;
        cap2 >> img_2;
        if (!img_1.data || !img_2.data)
        {
            cout << "error reading images " << endl;
            return -1;
        }
        //初始化
        clock_t startTime, endTime;
        startTime = clock();
        
        Ptr<ORB> orb = ORB::create(500, 1.2F, 8, 31, 0, 2, ORB::HARRIS_SCORE, 31, 20);//均为默认参数
        vector<KeyPoint> keyPoints_1, keyPoints_2;
        Mat descriptors_1, descriptors_2;

        //orb检测角点
        orb->detect(img_1, keyPoints_1);
        orb->detect(img_2, keyPoints_2);

        if (keyPoints_1.size() == 0 || keyPoints_2.size() == 0)
        {
            continue;
        }
        //计算描述子
        orb->compute(img_1, keyPoints_1, descriptors_1);
        orb->compute(img_2, keyPoints_2, descriptors_2);

        //匹配特征点，Hamming距离
        vector<DMatch> matches;
        BFMatcher matcher(NORM_HAMMING);
        matcher.match(descriptors_1, descriptors_2, matches);

        //筛选匹配点
        double min_dist = matches[0].distance, max_dist = matches[0].distance;

        for (int i = 0; i < descriptors_1.rows; i++)
        {
            double dist = matches[i].distance;
            if (dist < min_dist)
                min_dist = dist;
            if (dist > max_dist)
                max_dist = dist;
        }

        printf("max: %f\n", max_dist);
        printf("min: %f\n", min_dist);

        //当描述子之间的距离大于两倍的最小距离时,即认为匹配有误.但有时候最小距离会非常小,设置一个经验值30作为下限.
        std::vector< DMatch > good_matches;
        for (int i = 0; i < descriptors_1.rows; i++)
        {
            if (matches[i].distance <= max(2 * min_dist, 30.0))
            {
                good_matches.push_back(matches[i]);
            }
        }
        endTime = clock();
        cout << "Totle Time : " << (double)(endTime - startTime) / CLOCKS_PER_SEC << "s" << endl;
        printf("goodmatches number:%d\n", good_matches.size());
        //-- 第五步:绘制匹配结果
        /*Mat img_match;
        Mat img_goodmatch;
        drawMatches(img_1, keyPoints_1, img_2, keyPoints_2, matches, img_match);
        drawMatches(img_1, keyPoints_1, img_2, keyPoints_2, good_matches, img_goodmatch);
        imshow("所有匹配点对", img_match);
        imshow("优化后匹配点对", img_goodmatch);
        waitKey(1);*/
    }
    cap1.release();
    cap2.release();
    return 0;
}

测试效果

转载于:https://www.cnblogs.com/RegressionWorldLine/p/7471723.html