如何在 C++ 中获取边界框（矩形）内的像素值和坐标？有啥方法吗？

Posted 2023-03-27

【中文标题】如何在 C++ 中获取边界框（矩形）内的像素值和坐标？有啥方法吗？

【英文标题】：How can I get the pixels values and coordinates inside bounding box (rect) in c++? is there any method?如何在 C++ 中获取边界框（矩形）内的像素值和坐标？有什么方法吗？

【发布时间】：2021-11-01 09:44:15

【问题描述】：

我正在尝试在对象的边界框内获取像素的值及其 x 和 y 坐标。

这是我的部分代码：

 vector<Detector::Object> detected_objects;

 for (int i = 0; i < detected_objects.size(); ++i) 
    int xmin = detected_objects[i].rect.x;
    int ymin = detected_objects[i].rect.y;
    int width = detected_objects[i].rect.width;
    int height = detected_objects[i].rect.height;
    Rect rect(xmin, ymin, width, height); //The upper left coordinates (x, y) and the length (x) and width (y) of the rectangle
    cv::rectangle(osrc, rect, Scalar(200, 200, 10), 1, LINE_8, 0); // set rectangle color
   // std::cout << "\n coord: \n" << rect;
    //std::cout << "# of contour points: " << rect.size() << std::endl;
    int xmax = xmin + width;
    int ymax = ymin + height;
    for (size_t x = xmin; x < xmax; x++)
    
        for (size_t y = ymin; y < ymax; y++)
        

        
    

感谢任何帮助！

【问题讨论】：

创建 5 个像素作为示例。每边4个，中间1个。看看他们的坐标。看到模式了吗？

【参考方案1】：

v::Mat BoxValues = cv::Mat::zeros(detected_objects[i].rect.Height, detected_objects[i].rect.Width,type());
BoxValues = osrc(cv::Range(int(detected_objects[i].rect.y()),int(detected_objects[i].rect.y())+detected_objects[i].rect.Height),cv::Range(int(detected_objects[i].rect.x()),int(detected_objects[i].rect.x())+detected_objects[i].rect.Wigth));

结果 BoxValues 将保存边界框内像素值。 关于坐标希望我们可以通过cv::Range得到。

【讨论】：

谢谢！我试试看

【参考方案2】：

我找到了解决办法！

这是提取 yolo 边界框内的像素值和坐标的部分代码：

 for (i = 0; i < detected_objects.size(); ++i) 
    xmin = detected_objects[i].rect.x;
    ymin = detected_objects[i].rect.y;
    width = detected_objects[i].rect.width;
    height = detected_objects[i].rect.height;
    xmax = xmin + width;
    ymax = ymin + height;
    Rect rect(xmin, ymin, width, height); //The upper left coordinates (x, y) and the length (x) and width (y) of the rectangle
    cv::rectangle(osrc, rect, Scalar(200, 200, 10), 1, LINE_8, 0); // set rectangle color
    std::cout << "rect number:" << rect << "[" << i << "]" << "\n"; // print the number of detected objects

    for (x = xmin; x < xmax; x++)
    
        for (y = ymin; y < ymax; y++)
        
            // pixel_values = osrc.at<Vec3b>(x, y); // for 3 channels output
            // Scalar intensity = osrc.at<uchar>(y, x);
            Scalar intensity = osrc.at<uchar>(Point(x, y)); //get pixel intensity from one channel (Grayscale image)
            std::cout << "Pixel intensity: " << intensity <<  "[" << i << "]" << "\n";
            //allocate the array
            int** arr = new int* [xmax];
            for (int k = 0; k < xmax; k++)
                arr[k] = new int[ymax];

            // use the array
            cout << "coords: "<<"[" << x << "," << y << "]: \n ";
            //cout << arr[x][y] << endl;

            //deallocate the array
            for (int k = 0; k < xmax; k++)
                delete[] arr[k];
            delete[] arr;


            // std::cout << " \npixel: \n " << Point(x,y);
         //   pixelValue = GetPixel(hDC, pos.x, pos.y);
         //   std::cout << " \pixelValue: \n" << pixelValue;