smart-ocr

Smart OCR 技能

概述

本技能利用 PaddleOCR（支持100多种语言的领先OCR引擎）从图像和扫描文档中智能提取文本。可从照片、截图、扫描PDF和手写文档中高精度提取文本。

使用方法

提供图像或扫描文档
可选指定要检测的语言
我将提取文本并附带位置和置信度数据

示例提示：

"从这张截图中提取所有文本"
"OCR这个扫描的PDF文档"
"读取这张名片照片上的文字"
"从这张图片中提取中文和英文文本"

领域知识

PaddleOCR 基础

from paddleocr import PaddleOCR

# 初始化OCR引擎
ocr = PaddleOCR(use_angle_cls=True, lang='en')

# 对图像运行OCR
result = ocr.ocr('image.png', cls=True)

# 结果结构：[[box, (text, confidence)], ...]
for line in result[0]:
    box = line[0]      # [[x1,y1], [x2,y2], [x3,y3], [x4,y4]]
    text = line[1][0]  # 提取的文本
    conf = line[1][1]  # 置信度分数
    print(f"{text} ({conf:.2f})")

支持的语言

# 常用语言代码
languages = {
    'en': '英语',
    'ch': '中文（简体）',
    'cht': '中文（繁体）',
    'japan': '日语',
    'korean': '韩语',
    'french': '法语',
    'german': '德语',
    'spanish': '西班牙语',
    'russian': '俄语',
    'arabic': '阿拉伯语',
    'hindi': '印地语',
    'vi': '越南语',
    'th': '泰语',
    # ... 支持100多种语言
}

# 使用特定语言
ocr = PaddleOCR(lang='ch')  # 中文
ocr = PaddleOCR(lang='japan')  # 日语
ocr = PaddleOCR(lang='multilingual')  # 自动检测

配置选项

from paddleocr import PaddleOCR

ocr = PaddleOCR(
    # 检测设置
    det_model_dir=None,         # 自定义检测模型
    det_limit_side_len=960,     # 检测最大边长
    det_db_thresh=0.3,          # 二值化阈值
    det_db_box_thresh=0.5,      # 框分数阈值
    
    # 识别设置
    rec_model_dir=None,         # 自定义识别模型
    rec_char_dict_path=None,    # 自定义字符字典
    
    # 角度分类
    use_angle_cls=True,         # 启用角度分类
    cls_model_dir=None,         # 自定义分类模型
    
    # 语言
    lang='en',                  # 语言代码
    
    # 性能
    use_gpu=True,               # 如果可用则使用GPU
    gpu_mem=500,                # GPU内存限制（MB）
    enable_mkldnn=True,         # CPU优化
    
    # 输出
    show_log=False,             # 抑制日志
)

处理不同来源

图像文件

# 单张图像
result = ocr.ocr('image.png')

# 多张图像
images = ['img1.png', 'img2.png', 'img3.png']
for img in images:
    result = ocr.ocr(img)
    process_result(result)

PDF文件（扫描件）

from pdf2image import convert_from_path

def ocr_pdf(pdf_path):
    """OCR扫描的PDF。"""
    # 将PDF页面转换为图像
    images = convert_from_path(pdf_path)
    
    all_text = []
    for i, img in enumerate(images):
        # 保存临时图像
        temp_path = f'temp_page_{i}.png'
        img.save(temp_path)
        
        # OCR图像
        result = ocr.ocr(temp_path)
        
        # 提取文本
        page_text = '\n'.join([line[1][0] for line in result[0]])
        all_text.append(f"--- 第 {i+1} 页 ---\n{page_text}")
        
        os.remove(temp_path)
    
    return '\n\n'.join(all_text)

URL和字节数据

import requests
from io import BytesIO

# 从URL
response = requests.get('https://example.com/image.png')
result = ocr.ocr(BytesIO(response.content))

# 从字节数据
with open('image.png', 'rb') as f:
    img_bytes = f.read()
result = ocr.ocr(BytesIO(img_bytes))

结果处理

def process_ocr_result(result):
    """将OCR结果处理为结构化数据。"""
    
    lines = []
    for line in result[0]:
        box = line[0]
        text = line[1][0]
        confidence = line[1][1]
        
        # 计算边界框
        x_coords = [p[0] for p in box]
        y_coords = [p[1] for p in box]
        
        lines.append({
            'text': text,
            'confidence': confidence,
            'bbox': {
                'left': min(x_coords),
                'top': min(y_coords),
                'right': max(x_coords),
                'bottom': max(y_coords),
            },
            'raw_box': box
        })
    
    return lines

# 按位置排序（从上到下，从左到右）
def sort_by_position(lines):
    return sorted(lines, key=lambda x: (x['bbox']['top'], x['bbox']['left']))

文本布局重建

def reconstruct_layout(result, line_threshold=10):
    """从OCR结果重建文本布局。"""
    
    lines = process_ocr_result(result)
    lines = sort_by_position(lines)
    
    # 分组为逻辑行
    text_lines = []
    current_line = []
    current_y = None
    
    for line in lines:
        y = line['bbox']['top']
        
        if current_y is None or abs(y - current_y) < line_threshold:
            current_line.append(line)
            current_y = y
        else:
            # 新行
            text_lines.append(' '.join([l['text'] for l in current_line]))
            current_line = [line]
            current_y = y
    
    # 添加最后一行
    if current_line:
        text_lines.append(' '.join([l['text'] for l in current_line]))
    
    return '\n'.join(text_lines)

最佳实践

预处理图像：在OCR前提高质量
选择正确语言：指定语言以获得更高精度
处理多列：分别处理各列
过滤低置信度：跳过低于阈值的结果
批量处理：高效处理多张图像

常见模式

图像预处理

from PIL import Image, ImageEnhance, ImageFilter

def preprocess_image(image_path):
    """预处理图像以获得更好的OCR效果。"""
    img = Image.open(image_path)
    
    # 转换为灰度
    img = img.convert('L')
    
    # 增强对比度
    enhancer = ImageEnhance.Contrast(img)
    img = enhancer.enhance(2.0)
    
    # 锐化
    img = img.filter(ImageFilter.SHARPEN)
    
    # 保存预处理结果
    preprocessed_path = 'preprocessed.png'
    img.save(preprocessed_path)
    
    return preprocessed_path

批量OCR带进度

from tqdm import tqdm
from concurrent.futures import ThreadPoolExecutor

def batch_ocr(image_paths, max_workers=4):
    """并行OCR多张图像。"""
    
    results = {}
    
    def process_single(img_path):
        result = ocr.ocr(img_path)
        return img_path, result
    
    with ThreadPoolExecutor(max_workers=max_workers) as executor:
        futures = [executor.submit(process_single, p) for p in image_paths]
        
        for future in tqdm(futures, desc="处理OCR"):
            path, result = future.result()
            results[path] = result
    
    return results

示例

示例1：名片读取器

from paddleocr import PaddleOCR
import re

def read_business_card(image_path):
    """从名片中提取联系信息。"""
    
    ocr = PaddleOCR(use_angle_cls=True, lang='en')
    result = ocr.ocr(image_path)
    
    # 提取所有文本
    all_text = []
    for line in result[0]:
        all_text.append(line[1][0])
    
    full_text = '\n'.join(all_text)
    
    # 解析联系信息
    contact = {
        'name': None,
        'email': None,
        'phone': None,
        'company': None,
        'title': None,
        'raw_text': full_text
    }
    
    # 电子邮件模式
    email_match = re.search(r'[\w\.-]+@[\w\.-]+\.\w+', full_text)
    if email_match:
        contact['email'] = email_match.group()
    
    # 电话模式
    phone_match = re.search(r'[\+\d][\d\s\-\(\)]{8,}', full_text)
    if phone_match:
        contact['phone'] = phone_match.group().strip()
    
    # 姓名通常是最大/第一个文本
    if all_text:
        contact['name'] = all_text[0]
    
    return contact

card_info = read_business_card('business_card.jpg')
print(f"姓名: {card_info['name']}")
print(f"邮箱: {card_info['email']}")
print(f"电话: {card_info['phone']}")

示例2：收据扫描器

from paddleocr import PaddleOCR
import re

def scan_receipt(image_path):
    """从收据中提取项目和总金额。"""
    
    ocr = PaddleOCR(use_angle_cls=True, lang='en')
    result = ocr.ocr(image_path)
    
    lines = []
    for line in result[0]:
        text = line[1][0]
        y_pos = line[0][0][1]
        lines.append({'text': text, 'y': y_pos})
    
    # 按垂直位置排序
    lines.sort(key=lambda x: x['y'])
    
    receipt = {
        'items': [],
        'subtotal': None,
        'tax': None,
        'total': None
    }
    
    for line in lines:
        text = line['text']
        
        # 查找总金额
        if 'total' in text.lower():
            amount = re.search(r'\$?([\d,]+\.?\d*)', text)
            if amount:
                if 'sub' in text.lower():
                    receipt['subtotal'] = float(amount.group(1).replace(',', ''))
                else:
                    receipt['total'] = float(amount.group(1).replace(',', ''))
        
        # 查找税费
        elif 'tax' in text.lower():
            amount = re.search(r'\$?([\d,]+\.?\d*)', text)
            if amount:
                receipt['tax'] = float(amount.group(1).replace(',', ''))
        
        # 查找项目（带价格的行）
        else:
            item_match = re.search(r'(.+?)\s+\$?([\d,]+\.?\d+)$', text)
            if item_match:
                receipt['items'].append({
                    'name': item_match.group(1).strip(),
                    'price': float(item_match.group(2).replace(',', ''))
                })
    
    return receipt

receipt_data = scan_receipt('receipt.jpg')
print(f"项目数: {len(receipt_data['items'])}")
print(f"总金额: ${receipt_data['total']}")

示例3：多语言文档

from paddleocr import PaddleOCR

def ocr_multilingual(image_path, languages=['en', 'ch']):
    """OCR包含多种语言的文档。"""
    
    all_results = {}
    
    for lang in languages:
        ocr = PaddleOCR(use_angle_cls=True, lang=lang)
        result = ocr.ocr(image_path)
        
        texts = []
        for line in result[0]:
            texts.append({
                'text': line[1][0],
                'confidence': line[1][1]
            })
        
        all_results[lang] = texts
    
    # 合并结果，保留最高置信度
    merged = {}
    for lang, texts in all_results.items():
        for item in texts:
            text = item['text']
            conf = item['confidence']
            
            if text not in merged or merged[text]['confidence'] < conf:
                merged[text] = {'confidence': conf, 'language': lang}
    
    return merged

result = ocr_multilingual('bilingual_document.png')
for text, info in result.items():
    print(f"[{info['language']}] {text} ({info['confidence']:.2f})")

局限性

手写文本准确率因情况而异
非常小的文本可能无法检测
复杂背景会降低准确率
旋转文本需要角度分类
建议使用GPU以获得最佳性能

安装

# CPU版本
pip install paddlepaddle paddleocr

# GPU版本（CUDA 11.x）
pip install paddlepaddle-gpu paddleocr

# 额外依赖
pip install pdf2image Pillow

资源

你可能还喜欢

gemini

385Kmcp-integrations

Gemini CLI one-shot prompts, summaries, generation, skills, hooks, MCP, or Gemma routing.

steipete

获取

homelab-network-readiness

237Kmcp-integrations

Readiness checklist for homelab VLAN segmentation, local DNS filtering, and WireGuard-style remote access before changing router, firewall, DHCP, or VPN configuration.

affaan-m

获取

open-source

107Kmcp-integrations

Documentation reference for writing Python code using the browser-use open-source library. Use this skill whenever the user needs help with Agent, Browser, or Tools configuration, is writing code that imports from browser_use, asks about @sandbox deployment, supported LLM models, Actor API, custom tools, lifecycle hooks, MCP server setup, or monitoring/observability with Laminar or OpenLIT. Also trigger for questions about browser-use installation, prompting strategies, or sensitive data handling. Do NOT use this for Cloud API/SDK usage or pricing — use the cloud skill instead. Do NOT use this for directly automating a browser via CLI commands — use the browser-use skill instead.

browser-use

获取

继续探索

浏览 mcp-integrations 热门 Skills 最近更新