1621 字
8 分钟
invoice_ocr

#

最近实验室财务政策(报销的相关事情)有一些改变,需要自己去写入库单,然后去财务处报销,但是我不想手写Word,然后就有了这个工具。

之前财务也有用过一个OCR的工具,但是我觉得不太好用,所以就干脆自己写了一个。

在提高效率的同时,也可以学习一下OCR的相关知识。

用到的库#

语言当然不用说了,Python。

  • ocr(文字识别):百度的OCR api 增值税发票识别(封面也是这里找的)
  • json:用来解析百度OCR的返回值
  • python-docx:用来生成Word文档
  • pypdf2:用来生成PDF文档
  • datetime:用来生成时间戳

思路#

  1. 从淘宝等网站获取订单发票
  2. 用百度OCR识别发票信息
  3. 生成Word文档
  4. 将发票pdf重命名并归档

效果#

部分效果图如下:

代码#

很长且没有注释,但是我觉得很好理解,就不贴了。

其实自己也快忘了hhhh


from datetime import datetime
import shutil
from time import sleep
from aip import AipOcr
import PyPDF2 as pdf
import json
import docx
import os

""" 你的 APPID AK SK """
APP_ID = 'xxxx'
API_KEY = 'xxx'
SECRET_KEY = 'xxxx'

client = AipOcr(APP_ID, API_KEY, SECRET_KEY)


""" 读取pdf """


def get_file_content(filePath):
    with open(filePath, 'rb') as fp:
        return fp.read()


# 遍历文件夹下所有pdf文件
def get_pdffiles(dir_path):
    import os
    files = os.listdir(dir_path)
    pdf_files = []
    for file in files:
        if file.endswith('.pdf'):
            pdf_files.append(dir_path+'/'+file)
    return pdf_files


# 商品类
class Commodity:
    def __init__(self):
        self.name = ''
        self.type = ''
        self.unit = ''
        self.price = 0
        self.tax_rate = 0
        self.tax = 0
        self.total_price = 0
        self.num = 0

    

# 入库单类


# 入库单
# Warehouse receipt

class WarehouseReceipt():
    def __init__(self, pdf_file_path=''):
        self.order_num = ''
        self.order_date = ''
        self.supplier = ''
        self.commodity_list = []
        self.total_price = 0
        self.json = ''
        self.pdf_file_path = pdf_file_path
        self.company = ''
        self.c_price = ''

    def __str__(self):
        return '订单号:'+self.order_num+',订单日期:'+self.order_date+',供应商:'+self.supplier+',总价:'+self.total_price

    
    def ocr(self):
        # pdf_file_path = 'invoice/1.pdf'



        if self.get_pdf_num() > 1:
            #print('pdf文件页数大于1,暂不支持')
            return
        else:
            #print('pdf文件页数为1')
            #print('开始识别')
            #print(self.pdf_file_path)
            pdf_file = self.get_file_content()
            self.json = client.vatInvoicePdf(pdf_file)
            # with open('{}.json'.format(self.pdf_file_path), 'w', encoding='utf-8') as f:
            #     json.dump(self.json, f, ensure_ascii=False, indent=4)

            #print(self.json)


    def ocr_pic(self):
        # pdf 转图片
        def pdf_to_pic(self):
            from pdf2image import convert_from_path
            pages = convert_from_path(self.pdf_file_path, 200)
            for page in pages:
                page.save(self.pdf_file_path.replace('.pdf','.jpg'), 'JPEG')
        pdf_to_pic(self)
        # 识别图片
        image = self.get_pic_content()
        
        self.json = client.receipt(image)
        # 保存json数据
        with open('invoice/{}.json'.format(self.pdf_file_path), 'w', encoding='utf-8') as f:
            json.dump(self.json, f, ensure_ascii=False, indent=4)
        #print(self.json)


    def handle_json(self):

        json_data = self.json
        # 商品对象列表
        self.commodity_list = []

        words_result = json_data['words_result']
        for i in range(0, len(words_result['CommodityTaxRate'])):

            # 商品名称
            self.commodity_list.append(Commodity())
            if i == len(words_result['CommodityTaxRate'])-1:
                for j in range(int(words_result['CommodityTaxRate'][i]['row'])-1, len(words_result['CommodityName'])):
                    #print(i, j)
                    self.commodity_list[i].name += words_result['CommodityName'][j]['word']
                    # self.commodity_list[i]+=(words_result['CommodityName'][j]['word'])
            else:
                for j in range(int(words_result['CommodityTaxRate'][i]['row'])-1, int(words_result['CommodityTaxRate'][i+1]['row'])-1):
                    #print(i, j)
                    self.commodity_list[i].name += words_result['CommodityName'][j]['word']
                    # self.commodity_list[i]+=(words_result['CommodityName'][j]['word'])

            # 商品单价(含税)
            # 判断是否含税
            if '%' in words_result['CommodityTaxRate'][i]['word']:
                #print('含税',i)
                #print('ll',self.commodity_list)
                
                #print('ll',self.commodity_list[i])

                self.commodity_list[i].price = float(
                    words_result['CommodityPrice'][i]['word']) + float(words_result['CommodityTax'][i]['word']) / int(words_result['CommodityNum'][i]['word'])
                #print('%')
            else:
                self.commodity_list[i].price = float(
                    words_result['CommodityPrice'][i]['word'])

            # 商品类型(只能单页)
            if words_result['CommodityType']:
                self.commodity_list[i].type = words_result['CommodityType'][i]['word']

            #self.commodity_list[i].price = words_result['CommodityPrice'][i]['word']

            # 单个商品总价
            self.commodity_list[i].total_price = self.commodity_list[i].price * \
                int(words_result['CommodityNum'][i]['word'])

            # 处理单价和总价的小数点
            self.commodity_list[i].price = round(
                self.commodity_list[i].price, 2)
            self.commodity_list[i].total_price = round(
                self.commodity_list[i].total_price, 2)

            self.commodity_list[i].num = int(
                words_result['CommodityNum'][i]['word'])

            # 商品单位
            self.commodity_list[i].unit = words_result['CommodityUnit'][i]['word']

            
        self.total_price = float(words_result['AmountInFiguers'])
        self.order_date = words_result['InvoiceDate']
        self.company = words_result['SellerName']
        self.c_price = words_result['AmountInWords']
            


        

            
            

            # 商品名称

        #print(self.commodity_list)
        for i in self.commodity_list:
            #print(i.name)
            #print(i.price)
            #print(i.total_price)
            #print(i.num)
            #print(i.type)
            #print(i.unit)
            pass

    # 生成入库单word文件

    def generate_word(self):
        from docx.shared import Pt
        from docx.oxml.ns import qn
        from docx.shared import Inches
        
        from docx.enum.text import WD_ALIGN_PARAGRAPH
        import datetime
        from docx.enum.table import WD_TABLE_ALIGNMENT
        from docx.enum.table import WD_ALIGN_VERTICAL
        from docx.enum.text import WD_PARAGRAPH_ALIGNMENT


        # 生成word文件
        doc = docx.Document()
        # 设置所有字体为宋体
        doc.styles['Normal'].font.name = u'宋体'


        doc.styles['Normal']._element.rPr.rFonts.set(qn('w:eastAsia'), u'宋体')
        # 设置所有段落的行距为1.5倍
        doc.styles['Normal'].paragraph_format.line_spacing = 1.5
        # 设置
        title = doc.add_paragraph("浙江工业大学耗材入库单", style='Normal')
        # 设置标题字体大小
        title.runs[0].font.size = Pt(18)
        # 加粗
        title.runs[0].font.bold = True
        # 设置标题居中
        title.paragraph_format.alignment = WD_ALIGN_PARAGRAPH.CENTER

        # 添加当天时间
        today = datetime.datetime.now().strftime('%Y年%m月%d日')
        today = doc.add_paragraph(today, style='Normal')
        today.paragraph_format.alignment = WD_ALIGN_PARAGRAPH.CENTER

        # 添加部门公章
        doc.add_paragraph("部门(公章):							单位:	元", style='Normal')




        # 设置表格
        table = doc.add_table(rows=1, cols=7, style='Table Grid')
        # 左下
        table.alignment = WD_ALIGN_VERTICAL.CENTER
        # table.alignment = 
        # table.alignment = WD_TABLE_ALIGNMENT.LEFT
        # table.alignment = WD_TABLE_ALIGNMENT.BOTTOM

        # 根据内容自动调整列宽
        # table.autofit = True
        # table.autofitcontent = True
        # 设置每一列的宽度
        table.cell(0, 1).width = Inches(4)
        table.cell(0, 2).width = Inches(3)
        

        # 设置 首行
        table.cell(0, 0).text = '供货单位'
        table.cell(0, 1).text = self.company
        table.cell(0, 2).text = '入库日期'
        table.cell(0, 3).text = self.order_date
        # 合并单元格
        table.cell(0, 3).merge(table.cell(0, 4))
        table.cell(0, 5).text = '入库单号'
        table.cell(0, 6).text = self.order_num
        
        # 表格 根据内容自动调整
        table.allow_autofit = True


        hdr_cells = table.add_row().cells
        hdr_cells[0].text = '序号'
        hdr_cells[1].text = '耗材名称'
        hdr_cells[2].text = '规格型号'
        hdr_cells[3].text = '单位'
        hdr_cells[4].text = '数量'
        hdr_cells[5].text = '单价'
        hdr_cells[6].text = '金额'
        for index,i in enumerate(self.commodity_list):
            row_cells = table.add_row().cells
            row_cells[0].text = str(index+1)
            row_cells[1].text = i.name
            row_cells[2].text = i.type
            row_cells[3].text = i.unit
            row_cells[4].text = str(i.num)
            row_cells[5].text = str(i.price)
            row_cells[6].text = str(i.total_price)
        
        # 合计行
        row_cells = table.add_row().cells
        row_cells[0].text = '合计'
        # 合并单元格
        
        row_cells[1].text = '(大写)'
        row_cells[2].merge(row_cells[3]).merge(row_cells[4]).text = self.c_price
        row_cells[5].text = '(小写)'
        row_cells[6].text = str(self.total_price)

        

        # 添加制表等
        doc.add_paragraph("制表(保管):			采购:			入库验收:", style='Normal')

        last = doc.add_paragraph(
            "第一联:存根联	第二联:财务	第三联:供货单位(采购)	第四联:验收人", style='Normal')
        last.runs[0].font.size = Pt(10)
        last.paragraph_format.alignment = WD_ALIGN_PARAGRAPH.CENTER
        

        # 保存文件

        doc.save('invoice/word/'+str(self.total_price)+self.company+self.order_date+'.docx')
    

    def get_file_content(self):
        with open(self.pdf_file_path, 'rb') as fp:
            return fp.read()

    def get_pic_content(self):
        with open(self.pdf_file_path.replace('.pdf','.jpg'), 'rb') as fp:
            return fp.read()

    # 获取pdf页码

    def get_pdf_num(self):
        pdfFileObj = open(self.pdf_file_path, 'rb')
        pdfReader = pdf.PdfFileReader(pdfFileObj)
        num = pdfReader.numPages
        return num

    def cp(self):
        # 复制并重命名pdf文件
        shutil.copy(self.pdf_file_path, 'invoice/pdf/'+str(self.total_price)+self.company+self.order_date+'.pdf')
        # 删除原pdf文件
        os.remove(self.pdf_file_path)

    def deal(self):
        self.ocr()
        # 捕捉异常
        try:
            self.handle_json()
        except Exception as e:
            print(e, 'json文件处理失败')
            # 创建错误日志
            with open('invoice/log/error-{}.log'.format(datetime.now().strftime('%Y年%m月%d日%H-%M')), 'a', encoding='utf-8') as f:
                f.write(self.pdf_file_path+'  '+str(e) +
                        '\r\n'+str(self.json)+'\r\n')
            #print(e)
            #print('处理json文件出错')
        # 抛出异常
        else:
            #print('处理json文件成功')
            print('处理json文件成功',self.pdf_file_path)
            # 创建成功日志
            with open('invoice/log/success-{}.log'.format(datetime.now().strftime('%Y年%m月%d日%H-%M')), 'a', encoding='utf-8') as f:
                f.write(self.pdf_file_path+'\r\n')

        self.generate_word()
        self.cp()
        # options = {}
        # options['pdf_file_num'] = get_pdf_num(pdf_file_path)
        # res_pdf = client.vatInvoicePdf(pdf_file, options)
        # # 保存json数据
        # with open('invoice/1.json', 'w', encoding='utf-8') as f:
        #     json.dump(res_pdf, f, ensure_ascii=False, indent=4)

    def dealdebug(self):
        self.ocr()
        # 捕捉异常
        
        self.handle_json()
        

        self.generate_word()
        # options = {}
        # options['pdf_file_num'] = get_pdf_num(pdf_file_path)
        # res_pdf = client.vatInvoicePdf(pdf_file, options)
        # # 保存json数据
        # with open('invoice/1.json', 'w', encoding='utf-8') as f:
        #     json.dump(res_pdf, f, ensure_ascii=False, indent=4)

# read json


# def read_json(file_path):
#     with open(file_path, 'r', encoding='utf-8') as f:
#         json_data = json.load(f)
#         handle_json(json_data)
if __name__ == '__main__':

    for i in get_pdffiles('invoice/new'):
        WarehouseReceipt(i).deal()
    
    #WarehouseReceipt('invoice/new/202205171971483062.pdf').dealdebug()


    # WarehouseReceipt('invoice\new\202209102153729425.pdf').deal()












    # wh1 = WarehouseReceipt('invoice/invoice.pdf')

    # with open('invoice/invoice.json', 'r', encoding='utf-8') as f:
    #     wh1.json = json.load(f)
    # wh1.ocr()
    # wh1.handle_json()
    # wh1.deal()
    # wh1.handle_json()

    # wh1 = WarehouseReceipt()
    # with open('invoice/invoice.json', 'r', encoding='utf-8') as f:
    #     wh1.json = json.load(f)
    # wh1.handle_json()
    # wh1.generate_word()

    # read_json('invoice/invoice.json')
    # read_json('invoice/1.json')

    # pdf_file_path = 'invoice/1.pdf'
    # pdf_file = get_file_content(pdf_file_path)
    # options = {}
    # options['pdf_file_num'] = get_pdf_num(pdf_file_path)
    # res_pdf = client.vatInvoicePdf(pdf_file, options)
    # # 保存json数据
    # with open('invoice/1.json', 'w', encoding='utf-8') as f:
    #     json.dump(res_pdf, f, ensure_ascii=False, indent=4)


# 从文件读取json数据(测试用)


# 调用增值税发票识别
# res_image = client.vatInvoice(image)
# res_url = client.vatInvoiceUrl(url)
# res_pdf = client.vatInvoicePdf(pdf_file)
# #print(res_image)
# #print(res_url)
# #print(res_pdf)


# 如果有可选参数
# options = {}

# options['pdf_file_num'] = get_invoice_num(file_path)
# res_image = client.vatInvoice(image, options)
# res_url = client.vatInvoiceUrl(url, options)
# res_pdf = client.vatInvoicePdf(pdf_file, options)
# #print(res_image)
# #print(res_url)
# #print(res_pdf)
invoice_ocr
https://f.undf.top/posts/code/py/invoice_ocr/
作者
小魏的个人分享
发布于
2022-10-01
许可协议
CC BY-NC-SA 4.0