ocr_verify/merger/paddleocr_vl_merger.py

"""
PaddleOCR_VL 和 PaddleOCR 合并模块
"""
import json
from typing import List, Dict

try:
    from .text_matcher import TextMatcher
    from .bbox_extractor import BBoxExtractor
    from .data_processor import DataProcessor
    from .markdown_generator import MarkdownGenerator
except ImportError:
    from text_matcher import TextMatcher
    from bbox_extractor import BBoxExtractor
    from data_processor import DataProcessor
    from markdown_generator import MarkdownGenerator


class PaddleOCRVLMerger:
    """PaddleOCR_VL 和 PaddleOCR 结果合并器"""
    
    def __init__(self, look_ahead_window: int = 10, similarity_threshold: int = 90):
        """
        Args:
            look_ahead_window: 向前查找的窗口大小
            similarity_threshold: 文本相似度阈值
        """
        self.look_ahead_window = look_ahead_window
        self.similarity_threshold = similarity_threshold
        
        # 初始化子模块
        self.text_matcher = TextMatcher(similarity_threshold)
        self.bbox_extractor = BBoxExtractor()
        self.data_processor = DataProcessor(self.text_matcher, look_ahead_window)
        self.markdown_generator = MarkdownGenerator()
    
    def merge_table_with_bbox(self, paddleocr_vl_json_path: str, 
                             paddle_json_path: str) -> List[Dict]:
        """
        合并 PaddleOCR_VL 和 PaddleOCR 的结果
        
        Args:
            paddleocr_vl_json_path: PaddleOCR_VL 输出的 JSON 路径
            paddle_json_path: PaddleOCR 输出的 JSON 路径
        
        Returns:
            合并后的结果 (MinerU 格式)
        """
        # 加载数据
        with open(paddleocr_vl_json_path, 'r', encoding='utf-8') as f:
            paddleocr_vl_data = json.load(f)
        
        with open(paddle_json_path, 'r', encoding='utf-8') as f:
            paddle_data = json.load(f)
        
        # 提取 PaddleOCR 的文字框信息
        paddle_text_boxes = self.bbox_extractor.extract_paddle_text_boxes(paddle_data)
        
        # 处理 PaddleOCR_VL 的数据
        merged_data = self.data_processor.process_paddleocr_vl_data(
            paddleocr_vl_data, paddle_text_boxes
        )
        
        return merged_data
    
    def generate_enhanced_markdown(self, merged_data: List[Dict], 
                                   output_path: str = None,
                                   source_file: str = None) -> str:
        """生成增强的 Markdown"""
        return self.markdown_generator._generate_paddleocr_vl_markdown(
            merged_data, output_path, source_file
        )
    
    def extract_table_cells_with_bbox(self, merged_data: List[Dict]) -> List[Dict]:
        """提取所有表格单元格及其 bbox 信息"""
        return self.bbox_extractor.extract_table_cells_with_bbox(merged_data)