OpenCV4入门098：Hu矩实现轮廓匹配

索引地址：系列索引

OpenCV提供了一个名为matchShapes的函数，它接收两个图像（或轮廓）并使用Hu Moments找到它们之间的距离。所以，不必在自己计算hu矩。只需将图像二值化并使用matchShapes。

说明

函数比较两个形状，3个被实现的方法都使用hu 不变量。

声明

1
2
3
4
5
6
double matchShapes( 
       InputArray contour1,//第一张轮廓或者灰度图像
       InputArray contour2,//第二张轮廓或灰度图像
       int method,         //比较方法
       double parameter    //特定方法的参数，目前不支持
    );

method 参数

1
2
3
4
5
enum ShapeMatchModes {
        CONTOURS_MATCH_I1  =1, 
        CONTOURS_MATCH_I2  =2, 
        CONTOURS_MATCH_I3  =3  
    };

令 D(A,B)为形状A和B之间的距离，并且$H_i^A $和 HiBH_i^B为形状A和B的对数变换的Hu矩。定义对应于三种情况的距离:

1.CONTOURS_MATCH_I1

D(A,B)=∑i=06∣1HiB−1HiA∣D(A,B)=\sum_{i=0}^6 |\frac{1}{H_i^B}-\frac{1}{H_i^A}|

2.CONTOURS_MATCH_I2 =2

D(A,B)=∑i=06∣HiB−HiA∣D(A,B)=\sum_{i=0}^6 |H_i^B -H_i^A|

3.CONTOURS_MATCH_I3 =3

D(A,B)=∑i=06∣HiA−HiB∣∣HiA∣D(A,B)=\sum_{i=0}^6 \frac{|H_i^A-H_i^B|}{|H_i^A|}

如果两幅图一样那么距离为0。可以推测值越大两幅图越不像

测试代码：

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
#include <iostream>
#include <opencv2/opencv.hpp>

using namespace std;
using namespace cv;

void getHuMoments(Mat &src1,Mat &src2){
    //1.转化为灰度图像
    Mat gray1, binary1, dst1;
    Mat gray2, binary2, dst2;
    cvtColor(src1, gray1, COLOR_BGR2GRAY);
    cvtColor(src2, gray2, COLOR_BGR2GRAY);

    imshow("src1", src1);
    imshow("src2", src2);

    //2.二值化
    threshold(gray1, binary1, 127, 255, THRESH_BINARY | THRESH_OTSU);
    imshow("binary1", binary1);
    threshold(gray2, binary2, 127, 255, THRESH_BINARY | THRESH_OTSU);
    imshow("binary2", binary2);

    //3.计算hu矩
    Moments ms1,ms2;
    ms1 = moments(binary1, false);
    ms2 = moments(binary2, false);
    double hu1[7],hu2[7];
    HuMoments(ms1, hu1);
    HuMoments(ms2, hu2);

    //4.log转化
    cout << endl << endl;
    cout << "  img1:" << endl;
    for (size_t i = 0; i < 7; i++)
    {
        hu1[i] = -1 * copysign(1.0, hu1[i]) * log10(abs(hu1[i]));
        cout << "    hu1[" << i+1 << "]=" << hu1[i] << endl;
    }
    cout << endl;

    cout << "  img2:" << endl;
    for (size_t i = 0; i < 7; i++)
    {
        hu2[i] = -1 * copysign(1.0, hu2[i]) * log10(abs(hu2[i]));
        cout << "    hu2[" << i+1 << "]=" << hu2[i] << endl;
    }

    //5.形状匹配
    double d1 = matchShapes(binary1, binary2, CONTOURS_MATCH_I1, 0);
    double d2 = matchShapes(binary1, binary2, CONTOURS_MATCH_I2, 0);
    double d3 = matchShapes(binary1, binary2, CONTOURS_MATCH_I3, 0);

    cout << "  hu矩匹配值：" << endl;
    cout << "    CONTOURS_MATCH_T1 :" << d1 << endl;
    cout << "    CONTOURS_MATCH_T2 :" << d2 << endl;
    cout << "    CONTOURS_MATCH_T3 :" << d3 << endl;
}

int main() {
    Mat src1 = imread("1.jpg");
    Mat src2 = imread("1.png");

    if (src1.empty()||src2.empty()) {
        cout << " input the image error!" << endl;
    }
    getHuMoments(src1, src2);

    waitKey(0);
    return 0;
}

测试图片1为tifa，测试图片2为从图片1中截取的部分脸

结果为：

humoments