OCR 4: từ dòng chữ đến thông tin có cấu trúc

py -m venv .venv
.\.venv\Scripts\python.exe -m pip install opencv-python numpy pytesseract

from pathlib import Path
import shutil
import cv2
import numpy as np
import pytesseract

engine = shutil.which("tesseract")
if engine is None:
    candidate = Path(r"C:\Program Files\Tesseract-OCR\tesseract.exe")
    if candidate.is_file():
        engine = str(candidate)
if engine is None:
    raise RuntimeError("Install Tesseract or set its executable path first")
pytesseract.pytesseract.tesseract_cmd = engine

import re
import json

image = np.full((280, 1000, 3), 255, dtype=np.uint8)
for text, y in [("ORDER 2027", 110), ("TOTAL 125000", 215)]:
    cv2.putText(image, text, (35, y), cv2.FONT_HERSHEY_SIMPLEX,
                2.0, (0, 0, 0), 3, cv2.LINE_AA)
cv2.imwrite("receipt_demo.png", image)
rgb = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
raw = pytesseract.image_to_string(rgb, lang="eng", config="--psm 6")
Path("receipt_raw.txt").write_text(raw, encoding="utf-8")
print(repr(raw))

def parse_receipt(text):
    orders, totals = [], []
    for line in text.upper().splitlines():
        line = " ".join(line.split())
        order = re.fullmatch(r"ORDER\s+([0-9]{4})", line)
        total = re.fullmatch(r"TOTAL\s+([0-9]{1,9})", line)
        if order:
            orders.append(order.group(1))
        if total:
            totals.append(int(total.group(1)))
    valid = len(orders) == 1 and len(totals) == 1
    return {"status": "parsed" if valid else "needs_review",
            "order_id": orders[0] if valid else None,
            "total_vnd": totals[0] if valid else None}

result = parse_receipt(raw)
Path("receipt_fields.json").write_text(
    json.dumps(result, ensure_ascii=False, indent=2), encoding="utf-8")
print(result)

assert parse_receipt("ORDER 2027\nTOTAL 125000")["total_vnd"] == 125000
assert parse_receipt("ORDER 2027\nSUBTOTAL 125000")["status"] == "needs_review"
assert parse_receipt("ORDER 2027\nTOTAL 125000\nTOTAL 90000")["status"] == "needs_review"
assert parse_receipt("ORDER 2027\nTOTAL 12O000")["status"] == "needs_review"
print("Parser checks passed")

.\.venv\Scripts\python.exe ocr_receipt_fields.py