mescloud/_apk_work/extract_manual_images.py

# -*- coding: utf-8 -*-
"""Extract screenshots from 8.28 manuals with captions."""
import os, re, json
from pathlib import Path
import fitz

SRC = Path(r"D:\产线\大洋\大洋操作手册汇总8.28")
OUT = Path(r"D:\产线\大洋\大洋操作手册汇总9.19\images")
MIN_W, MIN_H = 180, 80
MIN_PIXELS = 25000

FASCICLES = [
    "1、系统管理操作手册.pdf",
    "2、设备管理操作手册.pdf",
    "3、物料管理操作手册.pdf",
    "4、质量管理操作手册.pdf",
    "5、生产管理操作手册.pdf",
    "6、ERP对接操作手册.pdf",
    "7、生产线管理操作手册.pdf",
    "8、PDA操作说明操作手册.pdf",
    "9、MES系统客户端员工扫码操作手册.pdf",
]


def caption_from_page(page, img_bbox):
    blocks = page.get_text("blocks")
    texts = []
    for b in blocks:
        x0, y0, x1, y1, text, *_ = b[:6]
        text = (text or "").strip()
        if not text:
            continue
        # caption usually below image, short
        texts.append((y0, y1, text.replace("\n", " ").strip()))
    # find nearest text below image
    below = [t for t in texts if t[0] >= img_bbox[3] - 8]
    below.sort(key=lambda t: t[0])
    for y0, y1, text in below[:3]:
        t = re.sub(r"\s+", " ", text)
        if 2 <= len(t) <= 40 and not t.startswith("第") and "PAGE" not in t:
            if not re.match(r"^\d+(\.\d+)*\.?\s*$", t):
                return t
    # fallback: last short line above
    above = [t for t in texts if t[1] <= img_bbox[1] + 8]
    if above:
        t = re.sub(r"\s+", " ", above[-1][2])
        if 2 <= len(t) <= 50:
            return t
    return ""


def heading_from_page(page):
    blocks = page.get_text("blocks")
    heads = []
    for b in blocks:
        text = (b[4] or "").strip()
        if not text:
            continue
        first = text.split("\n")[0].strip()
        if re.match(r"^\d+(\.\d+)*\.?\s+\S+", first) or first.startswith("3."):
            heads.append(first)
    return heads[0] if heads else ""


def safe_name(s):
    s = re.sub(r'[\\/:*?"<>|]', "", s)
    s = re.sub(r"\s+", "_", s.strip())
    return s[:40]


def extract_one(pdf_name):
    doc = fitz.open(SRC / pdf_name)
    stem = pdf_name[:-4]
    dest = OUT / stem
    dest.mkdir(parents=True, exist_ok=True)
    meta = []
    seen_xref = {}
    idx = 0
    for pi, page in enumerate(doc, 1):
        heading = heading_from_page(page)
        images = page.get_images(full=True)
        # also get image bboxes via dict
        img_info = []
        d = page.get_text("dict")
        for block in d.get("blocks", []):
            if block.get("type") != 1:
                continue
            bbox = block.get("bbox")
            w = bbox[2] - bbox[0]
            h = bbox[3] - bbox[1]
            img_info.append((bbox, w, h))
        # extract by xref, map to bbox by order
        for ii, img in enumerate(images):
            xref = img[0]
            try:
                pix = fitz.Pixmap(doc, xref)
            except Exception:
                continue
            if pix.n - pix.alpha > 3:
                pix = fitz.Pixmap(fitz.csRGB, pix)
            if pix.width < MIN_W or pix.height < MIN_H or pix.width * pix.height < MIN_PIXELS:
                continue
            # skip near-duplicates of same xref on same page
            key = (pi, xref, pix.width, pix.height)
            if key in seen_xref:
                continue
            seen_xref[key] = True
            bbox = img_info[ii][0] if ii < len(img_info) else (0, 0, pix.width, pix.height)
            cap = caption_from_page(page, bbox)
            idx += 1
            fname = "%03d_p%02d_%s.png" % (idx, pi, safe_name(cap) or "img")
            path = dest / fname
            pix.save(str(path))
            meta.append({
                "file": str(path.relative_to(OUT.parent)).replace("\\", "/"),
                "page": pi,
                "caption": cap,
                "heading": heading,
                "w": pix.width,
                "h": pix.height,
            })
    doc.close()
    (dest / "manifest.json").write_text(json.dumps(meta, ensure_ascii=False, indent=2), encoding="utf-8")
    print(stem, "images", len(meta))
    return meta


def main():
    OUT.mkdir(parents=True, exist_ok=True)
    allm = {}
    for name in FASCICLES:
        allm[name] = extract_one(name)
    (OUT / "all_manifest.json").write_text(json.dumps(allm, ensure_ascii=False, indent=2), encoding="utf-8")
    print("total", sum(len(v) for v in allm.values()))


if __name__ == "__main__":
    main()