Python实现发票信息识别

pip3 install fitz -i https://pypi.tuna.tsinghua.edu.cn/simplepip3 install PyMuPDF -i https://pypi.tuna.tsinghua.edu.cn/simplepip3 install paddleocr -i https://pypi.tuna.tsinghua.edu.cn/simplepip3 install paddlepaddle -i https://mirror.baidu.com/pypi/simple

import fitzfrom paddleocr import PaddleOCRfrom openpyxl import load_workbookpdf = fitz.open('test.pdf')page = pdf.load_page(0)  # 获取第0页# 提高分辨率pix = page.get_pixmap()zoom_x = 8zoom_y = 8mat = fitz.Matrix(zoom_x, zoom_y)pix = page.get_pixmap(matrix=mat, alpha=False)pix.save('temp.png')# for i in range(pdf.page_count):#     page = pdf.load_page(i)  # 获取第i页#     pix = page.get_pixmap()#     zoom_x = 8#     zoom_y = 8#     mat = fitz.Matrix(zoom_x, zoom_y)#     pix = page.get_pixmap(matrix=mat, alpha=False)#     pix.save('temp.png')ocr = PaddleOCR(use_angle_cls = True, use_gpu = False, lang='ch')res = ocr.ocr('temp.png', cls=True)invoice_dic = {}for r in res[0]:    text = r[1][0]    if text.startswith('发票号码：'):        invoice_dic.update({'number': text[5:]})        print(f"number：{text[5:]}")    if text.startswith('开票日期：'):        invoice_dic.update({'date': text[5:]})        print(f"date：{text[5:]}")    if text.startswith('名称：'):        if text.endswith('购买方名称'):            continue        invoice_dic.update({'name': text[3:]})        print(f"name：{text[3:]}")    if text.startswith('统一社会信用代码／纳税人识别号：'):        if text.endswith('购买方统一社会信用代码／纳税人识别号'):            continue        invoice_dic.update({'id': text[16:]})        print(f"id：{text[16:]}")    if text.startswith('（小写）¥'):        invoice_dic.update({'price': text[6:]})        print(f"price：{text[6:]}")

wb = load_workbook('test.xlsx') # 打开一个文件ws = wb.activenrow = ws.max_row # 取得行数append_row = nrow + 1ws.cell(append_row, 1).value = invoice_dic['number']ws.cell(append_row, 2).value = invoice_dic['date']ws.cell(append_row, 3).value = invoice_dic['name']ws.cell(append_row, 4).value = invoice_dic['id']ws.cell(append_row, 5).value = invoice_dic['price']wb.save('test.xlsx')

版权声明：本文内容来自CSDN:知数SEO，遵循CC 4.0 BY-SA版权协议上原文接及本声明。本作品采用知识共享署名-非商业性使用-禁止演绎 2.5 中国大陆许可协议进行可。原文链接：https://blog.csdn.net/weixin_42019349/article/details/140473184如有涉及到侵权，请联系，将立即予以删除处理。在此特别鸣谢原作者的创作。此篇文章的所有版权归原作者所有，与本公众号无关，商业转载建议请联系原作者，非商业转载请注明出处