限制tesseract OCR识别的字符数

Question

The images I'm trying to mask look like this. 我正在使用tesseract作为一个库来迭代多个图像，并识别其中的数字，然后对这些数字进行掩码。我在每个图像中有12个数字，我希望它在第8个数字之后停止识别/掩蔽，也就是说，只有8个数字需要被掩蔽，我尝试通过硬编码来实现这一点，因为它们中的大多数都有点相似，但是稍微偏离的图像会被搞乱。有没有办法阻止tesseract在8位数字之后屏蔽这些图像？我正在使用cv2.rectangle来遮罩图像。

pytesseract.pytesseract.tesseract_cmd = "D:/Tess/tesseract.exe"


for imgfilepathactual in glob.iglob('D:/dataset/allpdf/data/*.jpeg'):
    
    imgfilepath2 = imgfilepathactual.split("/")[3]
    imgfilepath1 = imgfilepath2.split('\\')[1]
    imgfilepath = imgfilepath1.split(".")[0]
    #print(filepath)
    print(imgfilepath)
    

 
            
    img = cv2.imread('D:/dataset/allpdf/data/' + imgfilepath + '.jpeg', cv2.IMREAD_GRAYSCALE) #if using with pdf conv
    #print(str(img))


    sobelX1 = cv2.Sobel(img, cv2.CV_64F, 1, 0, ksize = 1)
    sobelY1 = cv2.Sobel(img, cv2.CV_64F, 0, 1, ksize = 1)

    sobelX1 = np.uint8(np.absolute(sobelX1))
    sobelY1 = np.uint8(np.absolute(sobelY1))

    sobelCombined1 = cv2.bitwise_or(sobelX1, sobelY1)

    blurred = cv2.blur(sobelX1, (3,3)) #for pdf->img


    canny = cv2.Canny(blurred, 5, 250)

    pts1 =np.argwhere(canny>0)
    y11,x11 = pts1.min(axis=0)
    y21,x21 = pts1.max(axis=0)

    cropped = img[y11:y21, x11:x21]
    #cv2.imwrite("cropped.png", cropped)
    resizedimage = cv2.resize(cropped, (1080, 720), interpolation=cv2.INTER_CUBIC) #actual 
    cv2.imwrite('resizedimage' + imgfilepath + '.jpeg', resizedimage)

    img = cv2.imread('resizedimage' + imgfilepath + '.jpeg')
    h1, w1, _= img.shape

   

    resizedimage = Image.open('resizedimage' + imgfilepath + '.jpeg')
    box1 = (0,0, w1, 0.90*h1)

    resizedimage = resizedimage.crop(box1)
    resizedimage.save('resizedimage' + imgfilepath + '.jpeg')

    img = cv2.imread('resizedimage' + imgfilepath + '.jpeg')
    h2, w2, _ = img.shape

    print ((h2, w2), "reso")

    croppedimg2 = Image.open('resizedimage' + imgfilepath + '.jpeg')
    box2 = (0, 0.75*h2, 0.48*w2, h2)

    croppedimg2 = croppedimg2.crop(box2)
    croppedimg2.save('croppedimg2' + imgfilepath + '.jpeg')

    aadharBoxes = pytesseract.image_to_boxes(croppedimg2, lang = "eng")#, config=' --psm 7 -c tessedit_char_whitelist=0123456789abcdefghijklmnopqrstuvwxyz'
    #adharBoxes = aadharBoxes[0:8]
    for b in aadharBoxes.splitlines():
        b = b.split(' ')
        high = []
        for i in range(8):
            b[i] = b[i].split(' ')
            high.append(int(b[i][2]))
            high.append(int(b[i][4]))
        #b = b[0:8]
        print(b)
        #print(len(b[0]))
        maskedImage = cv2.rectangle(img, (int(b[0][1]), np.min(high)), (int(b[7][3]), np.max(high)), (0, 0, 0), -1)
        cv2.imwrite("maskedImage" + imgfilepath + ".jpeg", maskedImage)
        pdf = img2pdf.convert("maskedImage"+ imgfilepath + ".jpeg")
        file = open("D:/dataset/allpdf/masked_files/masked" + imgfilepath + ".pdf", "wb")
        file.write(pdf)
        file.close()
    #else:
        #pdf = img2pdf.convert("unmaskedImage"+ resizedimage + ".jpeg")
        #file = open("D:/dataset/allpdf/masked_files/masked" + imgfilepath + ".pdf", "wb")
        #file.write(pdf)
        #file.close()
        
    os.remove('resizedimage' + imgfilepath + '.jpeg')
    #os.remove('maskedImage' + imgfilepath + '.jpeg')
    #os.remove('croppedimg2' + imgfilepath + '.jpeg')
      


for filepathactual in glob.iglob('D:/dataset/allpdf/*.pdf'):
    #print(filepathactual)
    filepath2 = filepathactual.split("/")[2]
    filepath1 = filepath2.split("\\")[1]
    filepath = filepath1.split(".")[0]
    print(filepath)

    def convertPdf2img():
        pages = convert_from_path(filepathactual, 500)  #converting pdf to img
        for page in pages:
            page.save('out' + filepath + '.jpg', 'JPEG')


    
    convertPdf2img()
    


            
    img = cv2.imread('out' + filepath + '.jpg', cv2.IMREAD_GRAYSCALE) #if using with pdf conv



    sobelX1 = cv2.Sobel(img, cv2.CV_64F, 1, 0, ksize = 1)
    sobelY1 = cv2.Sobel(img, cv2.CV_64F, 0, 1, ksize = 1)

    sobelX1 = np.uint8(np.absolute(sobelX1))
    sobelY1 = np.uint8(np.absolute(sobelY1))

    sobelCombined1 = cv2.bitwise_or(sobelX1, sobelY1)

    blurred = cv2.blur(img, (3,3)) #for pdf->img


    canny = cv2.Canny(blurred, 5, 250)

    pts1 =np.argwhere(canny>0)
    y11,x11 = pts1.min(axis=0)
    y21,x21 = pts1.max(axis=0)

    cropped = img[y11:y21, x11:x21]
    #cv2.imwrite("cropped.png", cropped)
    resizedimage = cv2.resize(cropped, (1080, 720), interpolation=cv2.INTER_CUBIC) #actual 
    cv2.imwrite('resizedimage' +filepath + '.jpeg', resizedimage)

    img = cv2.imread('resizedimage' +filepath + '.jpeg')
    h1, w1, _= img.shape

   

    resizedimage = Image.open('resizedimage' +filepath + '.jpeg')
    box1 = (0,0, w1, 0.90*h1)

    resizedimage = resizedimage.crop(box1)
    resizedimage.save('resizedimage' +filepath + '.jpeg')

    img = cv2.imread('resizedimage' +filepath + '.jpeg')
    h2, w2, _ = img.shape

    print ((h2, w2), "reso")

    croppedimg2 = Image.open('resizedimage' +filepath + '.jpeg')
    box2 = (0, 0.65*h2, 0.6*w2, h2)

    croppedimg2 = croppedimg2.crop(box2)
    croppedimg2.save('croppedimg2' + filepath + '.jpeg')

    aadharBoxes = pytesseract.image_to_boxes(croppedimg2, lang = "eng")

    for b in aadharBoxes.splitlines():
        b = b.split(' ')
        maskedImage = cv2.rectangle(img, (int(b[1]), h2 - int(b[2])), (int(b[3]), h2 - int(b[4])), (0, 0, 0), -1)
        #print(b, "coords")
        cv2.imwrite("maskedImage" + filepath + ".jpeg", maskedImage)
        pdf = img2pdf.convert("maskedImage"+ filepath + ".jpeg")
        file = open("D:/dataset/allpdf/masked_files/masked" + filepath + ".pdf", "wb")
        file.write(pdf)
        file.close()
        


    #print(w,h)
    os.remove('out' + filepath + '.jpg')
    os.remove('resizedimage' + filepath + '.jpeg')
    os.remove('maskedImage' + filepath + '.jpeg')
    os.remove('croppedimg2' + filepath + '.jpeg')

Answer 1

要选择前八个字符，该范围之后的字符将被阻止。

import pytesseract
import numpy as np
import cv2

img = cv2.imread('muTYX.jpg')

pytesseract.pytesseract.tesseract_cmd = r'C:\Program Files\Tesseract-OCR\tesseract.exe'

text = pytesseract.image_to_boxes(img, lang = 'eng', config='--psm 7 --oem 3')

text = text.split('\n')
high = []
for i in range(8):
    text[i] = text[i].split(' ')
    high.append(int(text[i][2]))
    high.append(int(text[i][4]))

cv2.rectangle(img, (int(text[0][1]), np.min(high)), (int(text[7][3]), np.max(high)), (0, 255, 0), 2)
cv2.imshow('nubmer',img)
cv2.waitKey(0)

​

​

​

​