PDF转换成TXT脚本代码-QZQ-2026-9-28

# -*- coding: utf-8 -*- """ PDF转TXT工具 - 有文本层的PDF：直接用PyMuPDF提取 - 扫描件PDF：用RapidOCR识别（需要先 pip install rapidocr-onnxruntime pymupdf） 用法：python pdf2txt.py """ import pymupdf import os import time from rapidocr_onnxruntime import RapidOCR # ============ 配置 ============ SRC_DIR = r'F:\BaiduNetdiskDownload\python电子书籍' OUT_DIR = r'F:\BaiduNetdiskDownload\python电子书籍_TXT' # 要转换的文件（按需增删） PDF_FILES = [ 'Python算法教程_中文版.pdf', 'Python学习手册(第4版).pdf', ] ZOOM = 1.5 # 渲染分辨率，越大越准但越慢 # ============================== os.makedirs(OUT_DIR, exist_ok=True) engine = RapidOCR() def has_text_layer(doc, sample_pages=10): """检测PDF是否有文本层""" total = 0 for i in range(min(sample_pages, doc.page_count)): total += len(doc[i].get_text('text', sort=True).strip()) return total > 200 def extract_text_pdf(src_path, out_path): """直接提取文本层""" doc = pymupdf.open(src_path) total = doc.page_count with open(out_path, 'w', encoding='utf-8') as f: for i in range(total): text = doc[i].get_text('text', sort=True) f.write(f'\n===== 第 {i+1} 页 =====\n\n{text}\n') if (i+1) % 100 == 0: print(f' 提取中: {i+1}/{total}') doc.close() return total def ocr_scanned_pdf(src_path, out_path): """OCR识别扫描件，支持断点续传""" doc = pymupdf.open(src_path) total = doc.page_count # 断点续传：检查已完成页数 start_page = 0 mode = 'w' if os.path.exists(out_path): with open(out_path, 'r', encoding='utf-8') as f: done = f.read().count('===== 第 ') if done >= total: print(f' 已完成，跳过') doc.close() return total start_page = done mode = 'a' print(f' 从第 {start_page+1} 页继续') mat = pymupdf.Matrix(ZOOM, ZOOM) t0 = time.time() with open(out_path, mode, encoding='utf-8') as f: for i in range(start_page, total): pix = doc[i].get_pixmap(matrix=mat, alpha=False) tmp_img = os.path.join(OUT_DIR, f'_tmp_p{i}.png') pix.save(tmp_img) result, _ = engine(tmp_img) lines = [line[1] for line in result] if result else [] f.write(f'\n===== 第 {i+1} 页 =====\n\n' + '\n'.join(lines) + '\n') f.flush() os.remove(tmp_img) if (i+1) % 10 == 0: elapsed = time.time() - t0 speed = (i+1-start_page) / elapsed eta = (total-i-1) / speed / 60 print(f' OCR: {i+1}/{total} 页, 速度{speed:.1f}页/秒, 预计剩余{eta:.0f}分钟') doc.close() return total if __name__ == '__main__': for fname in PDF_FILES: src = os.path.join(SRC_DIR, fname) txt_name = os.path.splitext(fname)[0] + '.txt' out = os.path.join(OUT_DIR, txt_name) print(f'\n处理: {fname}') doc = pymupdf.open(src) pages = doc.page_count doc.close() print(f' 共 {pages} 页') if has_text_layer(pymupdf.open(src)): print(' -> 有文本层，直接提取') n = extract_text_pdf(src, out) else: print(' -> 扫描件，使用OCR识别') n = ocr_scanned_pdf(src, out) size = os.path.getsize(out) / 1024 print(f' 完成: {n}页, {size:.0f}KB -> {out}') print('\n全部完成！')