基于Pytesseract的驾驶证数据提取正则表达式问题与优化

优化说明 #

• 简化冗余正则逻辑，使用re.IGNORECASE参数省略大小写匹配规则
• 兼容OCR识别误差，包含拼写偏差（如Issue/Iseue、Validity/falidity）、特殊符号干扰等场景
• 新增签发日期、非营运有效期、营运有效期三个字段的提取逻辑
• 兼容两种常见OCR输出格式，包含样本中N:开头的姓名行场景

优化后完整代码 #

import pytesseract
import cv2
import re
import numpy as np
import datetime
from PIL import Image

def driver_license(filename):  
    """
    处理驾驶证图像OCR，提取核心信息
    """
    # 图像角度校正
    img = cv2.imread(filename)
    osd_info = pytesseract.image_to_osd(img)
    rotate_angle = int(re.search('(?<=Rotate: )\d+', osd_info).group(0))
    if rotate_angle != 0:
        pil_img = Image.open(filename)
        pil_img = pil_img.rotate(360 - rotate_angle, expand=True)
        pil_img.save(filename)
    
    # 图像预处理
    img = cv2.imread(filename)
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    kernel = np.ones((1, 1), np.uint8)
    processed_img = cv2.dilate(cv2.erode(gray, kernel, iterations=1), kernel, iterations=1)
    
    # OCR识别
    ocr_text = pytesseract.image_to_string(processed_img)
    lines = [line.strip() for line in ocr_text.split('\n') if line.strip()]
    
    # 初始化返回结果
    result = {
        'firstName': None,
        'lastName': None,
        'age': None,
        'documentNumber': None,
        'issue_date': None,
        'validity_nt': None,
        'validity_t': None
    }

    for idx, line in enumerate(lines):
        try:
            # 提取姓名：兼容Name开头和N开头两种格式
            if re.search(r'(name|^N\s*:)', line, re.IGNORECASE):
                clean_name = re.sub(r'[^a-zA-Z\s]', '', line)
                clean_name = re.sub(r'name', '', clean_name, flags=re.IGNORECASE).strip()
                name_parts = clean_name.split()
                if len(name_parts) >= 2:
                    result['firstName'] = ' '.join(name_parts[:-1])
                    result['lastName'] = name_parts[-1]
                # 兼容姓名在当前行下一行的场景
                elif not result['firstName'] and idx + 1 < len(lines):
                    next_line = re.sub(r'[^a-zA-Z\s]', '', lines[idx+1]).strip()
                    name_parts = next_line.split()
                    result['firstName'] = ' '.join(name_parts[:-1])
                    result['lastName'] = name_parts[-1]
            
            # 提取驾驶证编号：合并两种格式规则
            doc_match = re.search(r'[A-Z]{2}[-\s]?\d{13}', line, re.IGNORECASE)
            if doc_match:
                result['documentNumber'] = re.sub(r'[^A-Z0-9]', '', doc_match.group(), flags=re.IGNORECASE)
            
            # 提取年龄
            if re.search(r'DOB', line, re.IGNORECASE):
                year_match = re.search(r'\d{4}', line)
                if year_match:
                    result['age'] = datetime.datetime.now().year - int(year_match.group())
            
            # 提取签发日期：兼容拼写错误
            if re.search(r'I\w*e\s+Date', line, re.IGNORECASE):
                date_match = re.search(r'\d{2}/\d{2}/\d{4}', line)
                if date_match:
                    result['issue_date'] = date_match.group()
            
            # 提取有效期：兼容OCR识别误差
            if re.search(r'(validity|falidity)\s*\(NT\)', line, re.IGNORECASE):
                date_match = re.search(r'(\d{2}/\d{2}/\d{4}|\d{4})', line)
                if date_match:
                    result['validity_nt'] = date_match.group()
            if re.search(r'(validity|falidity)\s*\(T\)', line, re.IGNORECASE):
                val_content = line.split(':', 1)[-1].strip() if ':' in line else line
                result['validity_t'] = re.sub(r'[^A-Z0-9/]', '', val_content).strip()
        except Exception:
            continue

    return result

正则表达式简易学习指南 #

• 核心基础优先掌握：先记住常用元字符含义，\d匹配数字、\s匹配空白符、.匹配任意字符、*/+匹配重复次数、()用于分组提取需要的内容、[]用于匹配指定字符范围
• 常用参数掌握：re.IGNORECASE忽略大小写、re.MULTILINE多行匹配，两个参数可以覆盖绝大多数日常场景
• OCR场景匹配技巧：因为OCR存在识别误差，不用追求100%精准匹配关键词，比如匹配签发日期可以用I\w*e\s+Date，兼容中间字符识别错误的情况，优先提取后续符合格式的日期内容
• 练习方法：从固定格式的文本匹配练手，比如先练习匹配手机号、邮箱、日期，再逐步处理有识别误差的非标准化文本

内容的提问来源于stack exchange，提问作者senarijit1618