Question

我正在使用opencv-python计算两个图像之间的形状上下文距离。这是我的代码：

import cv2
import numpy as np
# images loaded as grayscale images
im1 = cv2.imread("first.bmp", 0)
im2 = cv2.imread("second.bmp", 0)

# converting to binary images
_, im1 = cv2.threshold(im1, 0, 255, cv2.THRESH_BINARY | cv2.THRESH_OTSU)
_, im2 = cv2.threshold(im2, 0, 255, cv2.THRESH_BINARY | cv2.THRESH_OTSU)

# finding contours!!
_, cntr1, _ = cv2.findContours(im1, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
_, cntr2, _ = cv2.findContours(im2, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)


# In the next few lines I combine different contours founded in each image 
# because the method that computes the distance needs single contour and 
# would gives an error if given a set of contours
temp = cntr1[0]
for i in range(1, len(cntr1)):
    temp = np.vstack([temp, cntr1[i]])

temp2 = cntr2[0]
for i in range(1, len(cntr2)):
    temp2 = np.vstack([temp2, cntr2[i]])


# creating shape context distance extractor object, using it to compute
# distance and printing result
scde = cv2.createShapeContextDistanceExtractor()
distance = scde.computeDistance(temp, temp2)
print(distance)

现在，只要两个图像相对相似，一切似乎就可以了。例如image a和image b的距离为0.2716，这是非常低的，因此我们可以得出结论，这些图像的形状相同。但是当图像的差异更大时，例如图像a和image c（水平翻转的图像），则会出现此错误：

cv2.error：OpenCV（3.4.2）C：\ projects \ opencv-python \ opencv \ modules \ core \ src \ matmul.cpp：1218：error：（-215：Assertion failed）type ==（（（函数'cv :: gemmImpl'中的（6）＆（（1 << 3）-1））+（（（（2）-1）<< 3））

此错误是什么意思？我该怎么解决？

顺便说一句，组合轮廓的代码对我提供的图片没有影响，因为所有图片都只有一个轮廓。您只需删除这些行并在倒数第二行中使用distance = scde.computeDistance(cntr1[0],cntr2[0])。

*真是个了不起的词

opencv形状上下文距离提取错误

0 个答案: