【OpenCV】图片几何特征提取

【OpenCV】图片几何特征提取

2016年08月16日 16:57:18 番茄发烧了 阅读数：673

版权声明：本文为博主原创文章，未经博主允许不得转载。 https://blog.csdn.net/bless2015/article/details/52222832

上一版不太好，又重新自己了一遍。

#include<cv.h>  
#include<highgui.h>  
using namespace cv;
using namespace std;
CvMemStorage *stroage;
IplImage *contourimage;
CvSeq *seq = NULL;
int ku = -1;
double pi = 3.141592;
int main()
{


    stroage = cvCreateMemStorage();
    Mat src = imread("E:\\14.jpg");
    Mat I;
    medianBlur(src, I, 3);
    //blur(src, I, Size(3, 3), Point(-1, -1));
    //GaussianBlur(src, I, Size(3, 3), 0, 0);
    //bilateralFilter(src, I, 5, 0.0, 10.0);
    contourimage = cvCreateImage(cvGetSize(&(CvMat)I), 8, 1);
    cvtColor(I, I, CV_BGR2GRAY);

    Mat contours;
    Canny(I, contours, 100, 255);
    threshold(contours, contours, 100, 255, CV_THRESH_TRUNC);
    int numcontours = cvFindContours(&(CvMat)contours, stroage, &seq, sizeof(CvContour), CV_RETR_LIST);

    CvMoments moments;
    CvHuMoments hu;
    cvMoments(&(CvMat)I, &moments, 0);
    cvGetHuMoments(&moments, &hu);

    if (ku != numcontours)
    {
        ku = numcontours;
        printf("contournum:::: %d \n", numcontours);
    }

    CvSeq *c = 0;
    int zz = 0;
    int totl = 0;

    cvSet(contourimage, cvScalar(255, 0, 255));
    cvSet(contourimage, cvScalar(125, 0, 125));
    CvPoint2D32f center;
    float radius;
    CvPoint2D32f rectpoint[4];

    CvContour *testcontour = 0;
    //c为轮廓顶点数组
    for (c = seq; c != NULL; c = c->h_next)
    {

        // 取得轮廓面积
        double testdbArea = fabs(cvContourArea(c, CV_WHOLE_SEQ));
        //取得轮廓长度
        double testdbLength = cvArcLength(c);

        c->block_max;

        if (testdbArea >= 50 && testdbLength <= 5000)
        {

            //点集的最外面（up-right）矩形边界
            CvRect testrect = cvBoundingRect(c);
            //轮廓最小外界矩形
            CvBox2D testbox = cvMinAreaRect2(c);
            //cvDrawContours(&(CvMat)contours, c, cvScalar(0, 255, 255), cvScalar(0, 255, 0), 0, 2);


            cvRectangle(&(CvMat)contours, cvPoint(testrect.x, testrect.y + testrect.height), cvPoint(testrect.x + testrect.width, testrect.y), cvScalar(255, 0, 255), 1);
            double width = testrect.width;
            double height = testrect.height;
            double juxingmianji = width*height;



            //找外界圆
            //cvMinEnclosingCircle(c, &center, &radius); 
            //cout <<  radius << endl;
            //cout << center.x << "；" << center.y << endl;
            //画外接圆
            //cvCircle(&(CvMat)contours, cvPointFrom32f(center), (int)radius, cvScalar(255, 0, 255), 2);

            //特征1矩形度
            float mianjibi = testdbArea / juxingmianji;
            //特征2延长度
            float changkuanbi = width / height;
            //特征3周长比
            float zhouchangbi = 2 * (height + width) / testdbLength;
            //特征4似圆度
            float r = 4 * pi*testdbArea / (testdbLength*testdbLength);
            //特征5形状复杂性
            float e = (testdbLength*testdbLength) / testdbArea;
            //不动点特征 hu.hu1    hu.hu2

            cout << "矩形度：" << mianjibi << endl;
            cout << "延长度：" << changkuanbi << endl;
            cout << "周长比：" << zhouchangbi << endl;
            cout << "似圆度：" << r << endl;
            cout << "形状复杂性：" << e << endl;
            cout << "一阶矩：" << hu.hu1 << endl;
            cout << "二阶矩：" << hu.hu2 << endl;

        }
    }
    namedWindow("Canny");
    imshow("Canny", contours);

    waitKey();
    return 0;
}