我正在嘗試比較兩個(gè)pdf。我的方法包括轉(zhuǎn)換兩個(gè) pdf 并比較相應(yīng)頁(yè)面的圖像。我最初編寫了一個(gè)簡(jiǎn)單的代碼，它比較與下面的第一個(gè)代碼相對(duì)應(yīng)的兩個(gè)圖像，該代碼可以正常工作。import cv2import numpy as nporiginal = cv2.imread("image_old/imageOld_1.jpg")image_to_compare = cv2.imread("image_new/imageNew_1.jpg")image1 = original.shapeimage2 = image_to_compare.shapeif original.shape == image_to_compare.shape: print("The images have same size and channels") difference = cv2.subtract(original, image_to_compare) r, g, b = cv2.split(difference) cv2.imshow("difference", cv2.resize( difference, None, fx=0.3, fy=0.3)) print(cv2.countNonZero(b)) if cv2.countNonZero(b) == 0 and cv2.countNonZero(g) == 0 and cv2.countNonZero(r) == 0: print("The images are completely Equal") else: print("The images are not equal")sift = cv2.xfeatures2d.SIFT_create()kp_1, desc_1 = sift.detectAndCompute(original, None)kp_2, desc_2 = sift.detectAndCompute(image_to_compare, None) print("Keypoints of 1st image: " + str(len(kp_1)))print("Keypoints of 2nd image: " + str(len(kp_2)))index_params = dict(algorithm=0, trees=5)search_params = dict()flann = cv2.FlannBasedMatcher(index_params, search_params)matches = flann.knnMatch(desc_1, desc_2, k=2)good_points = []for m, n in matches: if m.distance < 0.6*n.distance: good_points.append(m)print('The images have %d %s' %(len(good_points),"good points matches"))if len(kp_1) <= len(kp_2): number_keypoints = len(kp_1)else: number_keypoints = len(kp_2)percentage_similarity = len(good_points) / number_keypoints * 100print('Similarity %d %s' %(round((percentage_similarity)),"%\n"))result = cv2.drawMatches(original, kp_1, image_to_compare, kp_2, good_points, None)cv2.waitKey(0)cv2.destroyAllWindows()接下來(lái)，我添加了一個(gè) for 循環(huán)，與上面的代碼執(zhí)行相同的操作，但針對(duì) 5 個(gè)圖像。這個(gè)想法是，它獲取 imageOld_1.jpg 并與 imageNew_1.jpg、imageOld_2.jpg 進(jìn)行比較，并與 imageNew_2.jpg 進(jìn)行比較...下面的代碼比較第一對(duì)圖像 imageOld_1.jpg 和 imageNew_1.jpg（我得到結(jié)果）但由于某種原因，它不會(huì)繼續(xù)循環(huán)遍歷其他圖像，即比較 imageOld_2.jpg 和 imageNew_2.jpg、比較 imageOld_3.jpg 和 imageNew_3.jpg...我收到以下錯(cuò)誤，我做錯(cuò)了什么？
1 回答
明月笑刀無(wú)情
TA貢獻(xiàn)1828條經(jīng)驗(yàn) 獲得超4個(gè)贊
您需要組合兩個(gè)列表來(lái)比較相同的圖像。
您可以使用zip：
import cv2
a = [1]
b = [1]
for i, j in zip(a, b):
original = cv2.imread("image_old/imageOld_" + str(i) + ".jpg")
image_to_compare = cv2.imread("image_new/imageNew_" + str(j) + ".jpg")
image1 = original.shape
image2 = image_to_compare.shape
if original.shape == image_to_compare.shape:
print("The images " + str(i) + " have same size and channels")
print("Diffing page " + str(i) + " and " + str(j) + " of both pdfs")
difference = cv2.subtract(original, image_to_compare)
r, g, b = cv2.split(difference)
cv2.imshow("difference", cv2.resize(difference, None, fx=0.3, fy=0.3))
if cv2.countNonZero(b) == 0 and cv2.countNonZero(g) == 0 and cv2.countNonZero(r) == 0:
print("The images are completely Equal")
else:
print("The images are not equal")
sift = cv2.xfeatures2d.SIFT_create()
kp_1, desc_1 = sift.detectAndCompute(original, None)
kp_2, desc_2 = sift.detectAndCompute(image_to_compare, None)
print("Keypoints of 1st image: " + str(len(kp_1)))
print("Keypoints of 2nd image: " + str(len(kp_2)))
index_params = dict(algorithm=0, trees=5)
search_params = dict()
flann = cv2.FlannBasedMatcher(index_params, search_params)
matches = flann.knnMatch(desc_1, desc_2, k=2)
good_points = []
for m, n in matches:
if m.distance < 0.6 * n.distance:
good_points.append(m)
print('The images have %d %s' % (len(good_points), "good points matches"))
if len(kp_1) <= len(kp_2):
number_keypoints = len(kp_1)
else:
number_keypoints = len(kp_2)
percentage_similarity = len(good_points) / number_keypoints * 100
print('Similarity %d %s' % (round((percentage_similarity)), "%\n"))
result = cv2.drawMatches(original, kp_1, image_to_compare, kp_2, good_points, None)
cv2.waitKey(0)
cv2.destroyAllWindows()
結(jié)果：
The images 1 have same size and channels
Diffing page 1 and 1 of both pdfs
The images are completely Equal
Keypoints of 1st image: 8066
Keypoints of 2nd image: 8066
The images have 3117 good points matches
Similarity 39 %
imagesNew_1.jpg以下是我用于和的示例imagesOld_1.jpg
添加回答
舉報(bào)
0/150
提交
取消