如何在检测到面部opencv后展开矩形

Question

我编写代码，我希望如何在检测到面到人类的头部和颈部之后展开矩形

    #include <opencv2/imgproc/imgproc.hpp>  //this code detetced faces ///
    #include <opencv2/objdetect/objdetect.hpp>///////////
    #include <opencv2/highgui/highgui.hpp>
     ////////////////////
    using namespace cv;
    using namespace std;

    int main()
    {

        CascadeClassifier cascade;

        if (!cascade.load("haarcascade_frontalface_alt2.xml"))  //load harcascade xml
            return -1;

         Mat src = imread("11.jpg");  //read image 

        if (src.empty())
            return -1;
    cv::resize(src,src,cv::Size(600,600));  resize image 
        Mat gray;
        cvtColor(src, gray, CV_BGR2GRAY);
        equalizeHist(gray, gray);

        vector<Rect> faces;
        cascade.detectMultiScale(gray, faces, 1.2, 3,0,Size(30,30));

        for (size_t i = 0; i < faces.size(); i++)
        {
     /////////////////////////////
            Rect r = faces[i];
    //////////////////////////////       

              Mat faceROI = gray( faces[i] );
          int x = faces[i].x;
          int y = faces[i].y;
          int h =0.3*y+faces[i].height;
          int w = faces[i].width;
          printf("Orig dimensions after h * w crop 1: %dx%d\n",h,w);
            printf("Orig dimensions after  x* y crop 2: %dx%d\n",x,y);

           rectangle(src,Point (x,y),Point(x + w,y +h),Scalar(255,0,255),1,4,0);
             imshow("mmmmmmmmmmm.jpg",src );  //show image in windows 
        }
    ///////////////////////////////

        waitKey(0);

        return 0;
    }

Answer 1

要展开OpenCV面部矩形，您还可以使用以下运算符（1）到cv::Rect类

//expanding or shrinking a rectangle by a certain amount
cv::Rect += cv::Size
//shifting a rectangle by a certain offset
cv::Rect -= cv::Point

将面部矩形扩展10％的示例将是

cv::Size deltaSize( faces[i].width * 0.1f, faces[i].height * 0.1f ); // 0.1f = 10/100
cv::Point offset( deltaSize.width/2, deltaSize.height/2);
faces[i] += deltaSize;
faces[i] -= offset;

Answer 2

您需要调整高度并适当地开始y坐标。

简而言之，我觉得盒子​​的高度应该在两侧（顶部和底部）增加0.3倍。所以我从0.3*h中减去y并将0.3*2*h（即0.6 * h）添加到h。

这只是我的粗略计算。对于微调，您可以针对不同的面进行分析，并提出最适合您目的的测量。

以下是更改后的代码：

int x = faces[i].x;
int h_temp = faces[i].height;    // storing original height
int y = faces[i].y - h_temp*0.3; // y is reduced by 0.3*h
int w = faces[i].width;
int h = h_temp*1.6;              // height is increases by 60%

以下是我得到的结果。绿色显示原始我得到。蓝色是修改过的：

请记住，您想要选择多少，取决于您。调整以上值以获得所需的结果。