cv2.matchShape

        函数cv2.matchShape()可以比较两个形状或轮廓的相似度。如果返回值越小，匹配越好。它是根据Hu矩来计算的，Hu矩是归一化中心矩的线性组合，之所以这么做是为了能够获取代表图像的某个特征的矩函数，这些矩函数对某些变换如缩放，旋转，镜像映射（除了h1）具有不变性。

double cv::matchShapes	(	InputArray 	contour1,
InputArray 	contour2,
int 	method,
double 	parameter 
)		
Python:
cv.matchShapes(	contour1, contour2, method, parameter	) ->	retval

 比较一下上面这三张图：

In [4]: img1 = cv2.imread(r'E:\comprehensive_library\Intelligent_material_online\lib\retrieval\data\base\ornament_1.png
   ...: ',0)

In [9]: img2 = cv2.imread(r'E:\comprehensive_library\Intelligent_material_online\lib\retrieval\data\base\jd2.jpg',0)

In [10]: img3 = cv2.imread(r'E:\comprehensive_library\Intelligent_material_online\lib\retrieval\data\base\mizhi1.png',0
    ...: )

In [17]: ret,thresh1=cv2.threshold(img1,240,255,0)

In [18]: cv2.namedWindow('thresh1')
    ...: cv2.imshow('thresh1',thresh1)
    ...: cv2.waitKey(0)
    ...: cv2.destroyWindow('thresh1')
    ...:

In [19]: ret,thresh2=cv2.threshold(img2,240,255,0)

In [20]: cv2.namedWindow('thresh2')
    ...: cv2.imshow('thresh2',thresh2)
    ...: cv2.waitKey(0)
    ...: cv2.destroyWindow('thresh2')
    ...:

In [21]: ret,thresh3=cv2.threshold(img3,240,255,0)

In [22]: cv2.namedWindow('thresh3')
    ...: cv2.imshow('thresh3',thresh1)
    ...: cv2.waitKey(0)
    ...: cv2.destroyWindow('thresh3')
    ...:

In [24]: contours,hierarchy = cv2.findContours(thresh1,2,1)

In [25]: cnt1 = contours[0]

In [26]: cnt1
Out[26]: array([[[ 64, 137]]], dtype=int32)

In [27]: contours1,hierarchy = cv2.findContours(thresh2,2,1)

In [28]: contours3,hierarchy = cv2.findContours(thresh3,2,1)

In [29]: cnt2=contours1[0]

In [30]: cnt2
Out[30]: array([[[291, 303]]], dtype=int32)

In [31]: cnt3=contours3[0]

In [33]: ret = cv2.matchShapes(cnt1,cnt2,1,0.0)

In [34]: ret
Out[34]: 0.0

In [35]: ret = cv2.matchShapes(cnt1,cnt3,1,0.0)

In [36]: ret
Out[36]: 1.7976931348623157e+308

In [37]: ret = cv2.matchShapes(cnt1,cnt1,1,0.0)

In [38]: ret
Out[38]: 0.0

In [39]: ret = cv2.matchShapes(cnt3,cnt3,1,0.0)

In [40]: ret
Out[40]: 0.0