c++ - 在 OpenCV 中获取 Z "depth"的简单方法

Question

我需要从两个不同的相机检测一个蓝色物体和一个红色物体，现在需要的任务是定位每个物体在 3D 空间中的位置，这意味着对于每个物体我们必须有它的 x,y,z 坐标，我见过this video和 here女巫做的正是我想做的，但是第一个视频没有示例代码，我的代码现在看起来像这样，它让我得到 x，y 的红色/蓝色对象，但没有深度:

#include <iostream>
#include "opencv2/highgui/highgui.hpp"
#include "opencv2/imgproc/imgproc.hpp"
#include <opencv/highgui.h>

using namespace cv;
using namespace std;

int main(int argc, char** argv)
//int func(int argc, char** argv)
{
    VideoCapture cap(0); //capture the video from webcam
    VideoCapture cap1(1); //capture the video from extrenal camers
    if (!cap.isOpened())  // if not success, exit program
    {
        cout << "Cannot open the web cam" << endl;
        return -1;
    }

    if (!cap1.isOpened())  // if not success, exit program
    {
        cout << "Cannot open the External camera" << endl;
        return -1;
    }

    namedWindow("Control", CV_WINDOW_AUTOSIZE); //create a window called "Control"

    int iLowH = 170;
    int iHighH = 179;
    int iLowS = 150;
    int iHighS = 255;
    int iLowV = 60;
    int iHighV = 255;

    //Create trackbars in "Control" window to control the range of red detection
    createTrackbar("LowH", "Control", &iLowH, 179); //Hue (0 - 179)
    createTrackbar("HighH", "Control", &iHighH, 179);
    createTrackbar("LowS", "Control", &iLowS, 255); //Saturation (0 - 255)
    createTrackbar("HighS", "Control", &iHighS, 255);
    createTrackbar("LowV", "Control", &iLowV, 255);//Value (0 - 255)
    createTrackbar("HighV", "Control", &iHighV, 255);

    int iLastX = -1;    //last known co-ordinates of red object
    int iLastY = -1;
    int iLastX1 = -1;
    int iLastY1 = -1;

    //Capture a temporary image from both cameras to obtain size
    Mat imgTmp;
    cap.read(imgTmp);
    cap1.read(imgTmp);

    //Create a black image with the size as the camera output
    Mat imgLines = Mat::zeros(imgTmp.size(), CV_8UC3);;
    Mat imgLines1 = Mat::zeros(imgTmp.size(), CV_8UC3);;

    //loop of continuously capturing frames from video
    while (true)
    {
        Mat imgOriginal;

        Mat imgOriginal1;
        bool bSuccess = cap.read(imgOriginal); // read a new frame from video webcam
        bool bSuccess1 = cap1.read(imgOriginal1); // read a new frame from video external cam

        if (!bSuccess || !bSuccess1) //if not success, break loop
        {
            cout << "Cannot read a frame from video stream" << endl;
            break;
        }
        //WebCam code for image and tracking/detecting  
        Mat imgHSV;
        cvtColor(imgOriginal, imgHSV, COLOR_BGR2HSV); //Convert the captured frame from BGR to HSV to control range of color to obtain and be able to detect it
        Mat imgThresholded;
        inRange(imgHSV, Scalar(iLowH, iLowS, iLowV), Scalar(iHighH, iHighS, iHighV), imgThresholded); //Threshold the image at the colors within specified range
        //morphological opening (removes noise and similar colored objects appearing in thresholded image)
        erode(imgThresholded, imgThresholded, getStructuringElement(MORPH_ELLIPSE, Size(5, 5)));
        dilate(imgThresholded, imgThresholded, getStructuringElement(MORPH_ELLIPSE, Size(5, 5)));

        //morphological closing (removes noise appearing inside our object in the thresholded image)
        dilate(imgThresholded, imgThresholded, getStructuringElement(MORPH_ELLIPSE, Size(5, 5)));
        erode(imgThresholded, imgThresholded, getStructuringElement(MORPH_ELLIPSE, Size(5, 5)));
        //Calculate the moments of the thresholded image to calculate the object position
        Moments oMoments = moments(imgThresholded);
        double dM01 = oMoments.m01;
        double dM10 = oMoments.m10;
        double dArea = oMoments.m00;

        // if the area <= 10000, I consider that the there are no object in the image and it's because of the noise, the area is not zero 
        if (dArea > 10000)
        {
            //calculate the position of the ball
            int posX = dM10 / dArea;
            int posY = dM01 / dArea;

            if (iLastX >= 0 && iLastY >= 0 && posX >= 0 && posY >= 0)
            {
                //Draw a red line from the previous point to the current point
                line(imgLines, Point(posX, posY), Point(iLastX, iLastY), Scalar(0, 0, 255), 2);
            }

            iLastX = posX;  //current point becomes last known point and loop continues
            iLastY = posY;
        }
        imshow("Thresholded Image", imgThresholded); //show the thresholded image   
        imgOriginal = imgOriginal + imgLines;
        imshow("Original", imgOriginal); //show the original image with the tracking lines if exist
        //External Cam code track/detect
        Mat imgHSV1;
        cvtColor(imgOriginal1, imgHSV1, COLOR_BGR2HSV); //Convert the captured frame from BGR to HSV
        Mat imgThresholded1;
        inRange(imgHSV1, Scalar(iLowH, iLowS, iLowV), Scalar(iHighH, iHighS, iHighV), imgThresholded1); //Threshold the image
        //morphological opening (removes noise and similar colored objects appearing in thresholded image)
        erode(imgThresholded1, imgThresholded1, getStructuringElement(MORPH_ELLIPSE, Size(5, 5)));
        dilate(imgThresholded1, imgThresholded1, getStructuringElement(MORPH_ELLIPSE, Size(5, 5)));
        //morphological closing (removes noise appearing inside our object in the thresholded image)
        dilate(imgThresholded1, imgThresholded1, getStructuringElement(MORPH_ELLIPSE, Size(5, 5)));
        erode(imgThresholded1, imgThresholded1, getStructuringElement(MORPH_ELLIPSE, Size(5, 5)));

        //Calculate the moments of the thresholded image to calculate the object position
        Moments oMoments1 = moments(imgThresholded1);
        double dM011 = oMoments1.m01;
        double dM101 = oMoments1.m10;
        double dArea1 = oMoments1.m00;

        // if the area <= 10000, I consider that the there are no object in the image and it's because of the noise, the area is not zero 
        if (dArea1 > 10000)
        {
            //calculate the position of the ball
            int posX1 = dM101 / dArea1;
            int posY1 = dM011 / dArea1;

            if (iLastX1 >= 0 && iLastY1 >= 0 && posX1 >= 0 && posY1 >= 0)
            {
                //Draw a red line from the previous point to the current point
                line(imgLines1, Point(posX1, posY1), Point(iLastX1, iLastY1), Scalar(0, 0, 255), 2);
            }

            iLastX1 = posX1;
            iLastY1 = posY1;
        }

        imshow("Thresholded Image 2", imgThresholded1); //show the thresholded image
        imgOriginal1 = imgOriginal1 + imgLines1;
        imshow("Original 2", imgOriginal1); //show the original image

        if (waitKey(30) == 27) //wait for 'esc' key press for 30ms. If 'esc' key is pressed, break loop
        {
            cout << "esc key is pressed by user" << endl;
            break;
        }
    }
    return 0;
}

Answer 1

您的问题的答案是立体视觉。您需要对两个摄像机进行立体校准，以获得允许从 2 个 View 生成场景深度图的变换矩阵。 OpenCV 提供 some functions这样做。

Here是一个入门教程。