223 lines
8.8 KiB
Python
223 lines
8.8 KiB
Python
|
import os
|
|||
|
|
import sys
|
|||
|
|
from pathlib import Path
|
|||
|
|
from typing import Union
|
|||
|
|
import tempfile
|
|||
|
|
import shutil
|
|||
|
|
|
|||
|
|
class OCRService:
|
|||
|
|
"""OCR识别服务 - 支持 MinerU、百度云OCR API、PaddleOCR（生产模式，不支持演示）"""
|
|||
|
|
|
|||
|
|
def __init__(self):
|
|||
|
|
self.ocr_type = self._detect_ocr_type()
|
|||
|
|
self._initialize_ocr()
|
|||
|
|
|
|||
|
|
def _detect_ocr_type(self) -> str:
|
|||
|
|
"""检测可用的OCR类型"""
|
|||
|
|
# 最优先使用百度云OCR API（速度快、精度高、免费额度足够日常使用）
|
|||
|
|
if os.getenv("BAIDU_OCR_APP_ID") and os.getenv("BAIDU_OCR_API_KEY") and os.getenv("BAIDU_OCR_SECRET_KEY"):
|
|||
|
|
return "baidu_cloud"
|
|||
|
|
# 其次使用 MinerU（最强大的文档解析工具，但速度慢）
|
|||
|
|
elif self._check_mineru():
|
|||
|
|
return "mineru"
|
|||
|
|
# 再次使用PaddleOCR
|
|||
|
|
elif self._check_paddleocr():
|
|||
|
|
return "paddleocr"
|
|||
|
|
# 没有可用的OCR
|
|||
|
|
else:
|
|||
|
|
raise RuntimeError(
|
|||
|
|
"❌ 没有可用的OCR引擎！请至少配置以下一种：\n"
|
|||
|
|
"1. MinerU - 将 MinerU-master 文件夹放在桌面\n"
|
|||
|
|
"2. 百度OCR - 配置环境变量 BAIDU_OCR_*\n"
|
|||
|
|
"3. PaddleOCR - 运行 pip install paddleocr paddlepaddle"
|
|||
|
|
)
|
|||
|
|
|
|||
|
|
def _initialize_ocr(self):
|
|||
|
|
"""初始化OCR引擎"""
|
|||
|
|
if self.ocr_type == "mineru":
|
|||
|
|
try:
|
|||
|
|
# 添加 MinerU 路径
|
|||
|
|
mineru_path = Path(r"c:\Users\UI\Desktop\MinerU-master")
|
|||
|
|
if mineru_path.exists() and str(mineru_path) not in sys.path:
|
|||
|
|
sys.path.insert(0, str(mineru_path))
|
|||
|
|
|
|||
|
|
from demo.demo import parse_doc
|
|||
|
|
self.mineru_parse = parse_doc
|
|||
|
|
|
|||
|
|
try:
|
|||
|
|
from torch.serialization import add_safe_globals
|
|||
|
|
from doclayout_yolo.nn.tasks import YOLOv10DetectionModel
|
|||
|
|
add_safe_globals([YOLOv10DetectionModel])
|
|||
|
|
except Exception:
|
|||
|
|
pass
|
|||
|
|
|
|||
|
|
print("✓ 使用 MinerU 引擎（高精度文档解析）")
|
|||
|
|
except Exception as e:
|
|||
|
|
raise RuntimeError(f"❌ MinerU 初始化失败: {e}\n请安装完整依赖或使用其他OCR引擎")
|
|||
|
|
|
|||
|
|
elif self.ocr_type == "baidu_cloud":
|
|||
|
|
try:
|
|||
|
|
from aip import AipOcr
|
|||
|
|
app_id = os.getenv("BAIDU_OCR_APP_ID")
|
|||
|
|
api_key = os.getenv("BAIDU_OCR_API_KEY")
|
|||
|
|
secret_key = os.getenv("BAIDU_OCR_SECRET_KEY")
|
|||
|
|
self.baidu_client = AipOcr(app_id, api_key, secret_key)
|
|||
|
|
print("✓ 使用百度云OCR API（高精度）")
|
|||
|
|
except Exception as e:
|
|||
|
|
raise RuntimeError(f"❌ 百度云OCR初始化失败: {e}\n请检查环境变量配置")
|
|||
|
|
|
|||
|
|
elif self.ocr_type == "paddleocr":
|
|||
|
|
try:
|
|||
|
|
from paddleocr import PaddleOCR
|
|||
|
|
self.paddle_ocr = PaddleOCR(use_angle_cls=True, lang="ch", show_log=False)
|
|||
|
|
print("✓ 使用 PaddleOCR 引擎（本地离线）")
|
|||
|
|
except Exception as e:
|
|||
|
|
raise RuntimeError(f"❌ PaddleOCR 初始化失败: {e}\n请运行: pip install paddleocr paddlepaddle")
|
|||
|
|
|
|||
|
|
def _check_mineru(self) -> bool:
|
|||
|
|
"""检查MinerU是否可用"""
|
|||
|
|
try:
|
|||
|
|
mineru_path = Path(r"c:\Users\UI\Desktop\MinerU-master")
|
|||
|
|
return mineru_path.exists() and (mineru_path / "demo" / "demo.py").exists()
|
|||
|
|
except:
|
|||
|
|
return False
|
|||
|
|
|
|||
|
|
def _check_paddleocr(self) -> bool:
|
|||
|
|
"""检查PaddleOCR是否可用"""
|
|||
|
|
try:
|
|||
|
|
import paddleocr
|
|||
|
|
return True
|
|||
|
|
except ImportError:
|
|||
|
|
return False
|
|||
|
|
|
|||
|
|
def extract_text(self, file_path: Union[str, Path]) -> str:
|
|||
|
|
"""从图片或PDF中提取文本"""
|
|||
|
|
file_path = str(file_path)
|
|||
|
|
file_ext = Path(file_path).suffix.lower()
|
|||
|
|
|
|||
|
|
if file_ext == '.pdf':
|
|||
|
|
return self._extract_from_pdf(file_path)
|
|||
|
|
else:
|
|||
|
|
return self._extract_from_image(file_path)
|
|||
|
|
|
|||
|
|
def _extract_from_image(self, image_path: str) -> str:
|
|||
|
|
"""从图片中提取文本"""
|
|||
|
|
if self.ocr_type == "mineru":
|
|||
|
|
return self._extract_with_mineru(image_path)
|
|||
|
|
elif self.ocr_type == "baidu_cloud":
|
|||
|
|
return self._extract_with_baidu_cloud(image_path)
|
|||
|
|
elif self.ocr_type == "paddleocr":
|
|||
|
|
return self._extract_with_paddleocr(image_path)
|
|||
|
|
else:
|
|||
|
|
raise RuntimeError("OCR引擎未正确初始化")
|
|||
|
|
|
|||
|
|
def _extract_with_mineru(self, file_path: str) -> str:
|
|||
|
|
"""使用 MinerU 提取文本（支持PDF和图片）"""
|
|||
|
|
try:
|
|||
|
|
# 创建临时输出目录
|
|||
|
|
temp_dir = tempfile.mkdtemp(prefix="mineru_")
|
|||
|
|
|
|||
|
|
try:
|
|||
|
|
# 调用 MinerU 解析
|
|||
|
|
file_path_obj = Path(file_path)
|
|||
|
|
self.mineru_parse(
|
|||
|
|
path_list=[file_path_obj],
|
|||
|
|
output_dir=temp_dir,
|
|||
|
|
lang="ch", # 中文
|
|||
|
|
backend="pipeline", # 使用 pipeline 模式
|
|||
|
|
method="auto" # 自动检测
|
|||
|
|
)
|
|||
|
|
|
|||
|
|
# 读取生成的 markdown 文件
|
|||
|
|
md_files = list(Path(temp_dir).rglob("*.md"))
|
|||
|
|
if md_files:
|
|||
|
|
# 优先排除 layout / span / origin 等辅助文件
|
|||
|
|
content_files = [
|
|||
|
|
f for f in md_files
|
|||
|
|
if not any(x in f.stem for x in ['layout', 'span', 'origin'])
|
|||
|
|
]
|
|||
|
|
target_files = content_files or md_files
|
|||
|
|
with open(target_files[0], 'r', encoding='utf-8') as f:
|
|||
|
|
content = f.read()
|
|||
|
|
return content if content.strip() else "未识别到文本内容"
|
|||
|
|
|
|||
|
|
return "未识别到文本内容"
|
|||
|
|
|
|||
|
|
finally:
|
|||
|
|
# 清理临时目录
|
|||
|
|
try:
|
|||
|
|
shutil.rmtree(temp_dir)
|
|||
|
|
except:
|
|||
|
|
pass
|
|||
|
|
|
|||
|
|
except Exception as e:
|
|||
|
|
return f"MinerU识别出错: {str(e)}"
|
|||
|
|
|
|||
|
|
def _extract_with_baidu_cloud(self, image_path: str) -> str:
|
|||
|
|
"""使用百度云OCR API提取文本"""
|
|||
|
|
try:
|
|||
|
|
# 读取图片
|
|||
|
|
with open(image_path, 'rb') as f:
|
|||
|
|
image_data = f.read()
|
|||
|
|
|
|||
|
|
# 调用通用文字识别（高精度版）
|
|||
|
|
result = self.baidu_client.accurateBasic(image_data)
|
|||
|
|
|
|||
|
|
if 'error_code' in result:
|
|||
|
|
return f"百度OCR错误 ({result['error_code']}): {result.get('error_msg', '未知错误')}"
|
|||
|
|
|
|||
|
|
# 提取文本
|
|||
|
|
if 'words_result' in result:
|
|||
|
|
text_lines = [item['words'] for item in result['words_result']]
|
|||
|
|
return "\n".join(text_lines) if text_lines else "未识别到文本内容"
|
|||
|
|
|
|||
|
|
return "未识别到文本内容"
|
|||
|
|
|
|||
|
|
except Exception as e:
|
|||
|
|
return f"百度云OCR识别出错: {str(e)}"
|
|||
|
|
|
|||
|
|
def _extract_with_paddleocr(self, image_path: str) -> str:
|
|||
|
|
"""使用PaddleOCR提取文本"""
|
|||
|
|
try:
|
|||
|
|
result = self.paddle_ocr.ocr(image_path, cls=True)
|
|||
|
|
|
|||
|
|
if not result or not result[0]:
|
|||
|
|
return "未识别到文本内容"
|
|||
|
|
|
|||
|
|
# 提取所有文本行
|
|||
|
|
text_lines = []
|
|||
|
|
for line in result[0]:
|
|||
|
|
if line and len(line) >= 2:
|
|||
|
|
text_lines.append(line[1][0])
|
|||
|
|
|
|||
|
|
return "\n".join(text_lines) if text_lines else "未识别到文本内容"
|
|||
|
|
|
|||
|
|
except Exception as e:
|
|||
|
|
return f"PaddleOCR识别出错: {str(e)}"
|
|||
|
|
|
|||
|
|
def _extract_from_pdf(self, pdf_path: str) -> str:
|
|||
|
|
"""从PDF中提取文本"""
|
|||
|
|
# 优先使用 MinerU 处理 PDF（效果最好）
|
|||
|
|
if self.ocr_type == "mineru":
|
|||
|
|
return self._extract_with_mineru(pdf_path)
|
|||
|
|
|
|||
|
|
# 备选方案：使用 pdfplumber
|
|||
|
|
try:
|
|||
|
|
import pdfplumber
|
|||
|
|
|
|||
|
|
text_content = []
|
|||
|
|
with pdfplumber.open(pdf_path) as pdf:
|
|||
|
|
for page in pdf.pages:
|
|||
|
|
text = page.extract_text()
|
|||
|
|
if text:
|
|||
|
|
text_content.append(text)
|
|||
|
|
|
|||
|
|
return "\n\n".join(text_content) if text_content else "未提取到文本内容"
|
|||
|
|
|
|||
|
|
except ImportError:
|
|||
|
|
# PDF库不可用，尝试使用OCR处理PDF的图像
|
|||
|
|
return "PDF处理需要安装 pdfplumber 库\n可以运行: pip install pdfplumber"
|
|||
|
|
except Exception as e:
|
|||
|
|
return f"PDF处理出错: {str(e)}"
|
|||
|
|
|