神人答题卡填涂助手

# fill_sheet_auto_calibrate.py

import cv2

import numpy as np

from PIL import Image, ImageDraw, ImageFont

from pdf2image import convert_from_path

import img2pdf

import os

import textwrap

import sys

import io

import json

if sys.stdout.encoding != 'utf-8':

    sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding='utf-8')

SCRIPT_DIR = os.path.dirname(os.path.abspath(__file__))

POPPLER_PATH = os.path.join(SCRIPT_DIR, "poppler-25.12.0", "bin")

if not os.path.exists(POPPLER_PATH):

    for suffix in ["", "_x64", "-windows-x64"]:

        alt = os.path.join(SCRIPT_DIR, f"poppler-25.12.0{suffix}", "bin")

        if os.path.exists(alt):

            POPPLER_PATH = alt

            break

    else:

        raise FileNotFoundError("Poppler 路径不存在，请检查。")

PDF_INPUT = "1_2025-2026学年福州市高二年级适应性练习语文答题卡(1).pdf"

PDF_OUTPUT = "filled_语文答题卡.pdf"

COORDS_CACHE = os.path.join(SCRIPT_DIR, "coords_cache.json")

ANSWERS_JSON = os.path.join(SCRIPT_DIR, "answers.json")

# ================== 加载答案 ==================

def load_answers(json_path):

    with open(json_path, "r", encoding="utf-8") as f:

        raw = json.load(f)

    # 将键转为整数

    obj = {int(k): v for k, v in raw["objective"].items()}

    subj = {int(k): v for k, v in raw["subjective"].items()}

    return {"objective": obj, "subjective": subj}

try:

    ANSWERS = load_answers(ANSWERS_JSON)

except Exception as e:

    print(f"[ERROR] 无法加载答案文件 {ANSWERS_JSON}: {e}")

    sys.exit(1)

# ================== 自动校准模块 ==================

def calibrate_interactive(pdf_path, poppler_path):

    print("[INFO] 正在加载答题卡用于校准...")

    images = convert_from_path(pdf_path, dpi=400, poppler_path=poppler_path)

    img = np.array(images[0])

    h, w = img.shape[:2]

    MAX_DISPLAY_WIDTH = 1400

    scale = min(MAX_DISPLAY_WIDTH / w, 1.0)

    display_img = cv2.resize(img, (int(w * scale), int(h * scale)), interpolation=cv2.INTER_AREA)

    print("\n[INFO] 请按以下顺序点击（每点一个，终端会确认）：")

    objective_order = []

    for q in [1, 2, 6]:

        for opt in ["A", "B", "C", "D"]:

            objective_order.append((q, opt))

    for opt in ["A", "B", "C", "D", "E", "F", "G"]:

        objective_order.append((10, opt))

    for q in [11, 12, 15, 20, 21]:

        for opt in ["A", "B", "C", "D"]:

            objective_order.append((q, opt))

    # 移除了 23 的作文题

    subjective_order = [3, 4, 5, 7, 8, 9, 13, 14, 16, 17, 18, 19, 22]

    all_points = []

    current_index = 0

    def click_event(event, x, y, flags, param):

        nonlocal current_index

        if event == cv2.EVENT_LBUTTONDOWN and current_index < len(objective_order) + len(subjective_order):

            orig_x = int(round(x / scale))

            orig_y = int(round(y / scale))

            all_points.append((orig_x, orig_y))

            if current_index < len(objective_order):

                q, opt = objective_order[current_index]

                print(f"[SUCCESS] 已记录 第{q}题 {opt}: ({orig_x}, {orig_y})")

            else:

                idx = current_index - len(objective_order)

                q = subjective_order[idx]

                print(f"[SUCCESS] 已记录 主观题 {q} 起始点: ({orig_x}, {orig_y})")

            current_index += 1

            cv2.circle(display_img, (x, y), 5, (0, 255, 0), -1)

            cv2.imshow("[INFO] 校准模式 - 按顺序点击 (按 'q' 退出，'r' 重置)", display_img)

    cv2.imshow("[INFO] 校准模式 - 按顺序点击 (按 'q' 退出，'r' 重置)", display_img)

    cv2.setMouseCallback("[INFO] 校准模式 - 按顺序点击 (按 'q' 退出，'r' 重置)", click_event)

    total_clicks = len(objective_order) + len(subjective_order)

    print(f"\n[INFO] 共需点击定位 {total_clicks} 次")

    while True:

        key = cv2.waitKey(1) & 0xFF

        if key == ord('q'):

            break

        elif key == ord('r'):

            current_index = 0

            all_points.clear()

            display_img[:] = cv2.resize(img, (int(w * scale), int(h * scale)), interpolation=cv2.INTER_AREA)

            print("[WARNING] 已重置，重新开始点击")

            cv2.imshow("[INFO] 校准模式 - 按顺序点击 (按 'q' 退出，'r' 重置)", display_img)

    cv2.destroyAllWindows()

    if len(all_points) != total_clicks:

        print(f"[WARNING] 警告：只记录了 {len(all_points)} 个点，需要 {total_clicks} 个")

        return None, None

    # 构建 OBJECTIVE_BOXES

    obj_boxes = {}

    idx = 0

    for q in [1, 2, 6]:

        obj_boxes[q] = {opt: all_points[idx + i] for i, opt in enumerate(["A", "B", "C", "D"])}

        idx += 4

    obj_boxes[10] = {opt: all_points[idx + i] for i, opt in enumerate(["A", "B", "C", "D", "E", "F", "G"])}

    idx += 7

    for q in [11, 12, 15, 20, 21]:

        obj_boxes[q] = {opt: all_points[idx + i] for i, opt in enumerate(["A", "B", "C", "D"])}

        idx += 4

    # 构建 SUBJ_STARTS

    subj_starts = {}

    for i, q in enumerate(subjective_order):

        subj_starts[q] = all_points[idx + i]

    return obj_boxes, subj_starts

def save_coords(obj_boxes, subj_starts, path):

    data = {"objective": obj_boxes, "subjective_starts": subj_starts}

    with open(path, "w", encoding="utf-8") as f:

        json.dump(data, f, ensure_ascii=False, indent=2)

    print(f"[INFO] 坐标已保存至: {path}")

def load_coords(path):

    if not os.path.exists(path):

        return None, None

    with open(path, "r", encoding="utf-8") as f:

        data = json.load(f)

    # 键转为 int

    obj = {int(q): {opt: tuple(pos) for opt, pos in opts.items()} for q, opts in data["objective"].items()}

    subj = {int(q): tuple(pos) for q, pos in data["subjective_starts"].items()}

    return obj, subj

# ================== 渲染模块 ==================

COLUMN_WIDTH = 1400

FONT_PATH = os.path.join(SCRIPT_DIR, "QuanHengDuLiang-TTF.ttf")

FONT_SIZE = 70

LINE_SPACING = 134

FILL_RADIUS = 18

def build_subjective_boxes(subj_starts):

    boxes = {}

    for q, (x, y) in subj_starts.items():

        if q in [8, 9, 17]:

            boxes[q] = (x, y, COLUMN_WIDTH, 160)

        elif q in [5, 13, 14, 16, 19, 22]:

            boxes[q] = (x, y, COLUMN_WIDTH, 120)

        elif q == 3:

            boxes[q] = (x, y, COLUMN_WIDTH, 150)

        elif q == 4:

            boxes[q] = (x, y, COLUMN_WIDTH, 100)

        else:

            boxes[q] = (x, y, COLUMN_WIDTH, 100)

    return boxes

def pdf_to_image(pdf_path, poppler_path, dpi=400):

    images = convert_from_path(pdf_path, dpi=dpi, poppler_path=poppler_path)

    return np.array(images[0])

def fill_bubbles(img, answers, boxes, radius=18):

    for q, ans in answers.items():

        if q not in boxes:

            print(f"[WARNING] 客观题 {q} 未定义坐标")

            continue

        opts = boxes[q]

        if isinstance(ans, list):

            for a in ans:

                if a in opts:

                    x, y = opts[a]

                    cv2.circle(img, (x, y), radius, (0, 0, 0), -1)

        else:

            if ans in opts:

                x, y = opts[ans]

                cv2.circle(img, (x, y), radius, (0, 0, 0), -1)

def draw_text_in_box(draw, text, box, font, line_spacing):

    x, y, max_width, max_height = box

    paragraphs = text.split('\n')

    wrapped_lines = []

    avg_char_width = font.getlength("中")

    max_chars = max(1, int(max_width / avg_char_width))

    for para in paragraphs:

        if para.strip():

            lines = textwrap.wrap(para, width=max_chars, break_long_words=False)

            wrapped_lines.extend(lines)

        else:

            wrapped_lines.append("")

    total_height = len(wrapped_lines) * line_spacing

    start_y = y if total_height > max_height else y + (max_height - total_height) // 2

    for i, line in enumerate(wrapped_lines):

        draw.text((x, start_y + i * line_spacing), line, fill=(0, 0, 0), font=font)

def draw_text_on_image(pil_img, answers, boxes, font_path, font_size, line_spacing):

    try:

        font = ImageFont.truetype(font_path, font_size)

    except:

        print(f"[WARNING] 无法加载字体 {font_path}")

        font = ImageFont.load_default()

    draw = ImageDraw.Draw(pil_img)

    for q, text in answers.items():

        if q not in boxes:

            print(f"[WARNING] 主观题 {q} 无定义区域")

            continue

        draw_text_in_box(draw, text, boxes[q], font, line_spacing)

# ================== 主流程 ==================

def main():

    obj_boxes, subj_starts = load_coords(COORDS_CACHE)

    if obj_boxes is None or subj_starts is None:

        print("[INFO] 未找到坐标缓存，进入校准模式...")

        obj_boxes, subj_starts = calibrate_interactive(PDF_INPUT, POPPLER_PATH)

        if obj_boxes is None:

            print("[ERROR] 校准未完成，退出。")

            return

        save_coords(obj_boxes, subj_starts, COORDS_CACHE)

    else:

        print(f"[INFO] 已从 {COORDS_CACHE} 加载坐标")

    img = pdf_to_image(PDF_INPUT, POPPLER_PATH, dpi=400)

    fill_bubbles(img, ANSWERS["objective"], obj_boxes, FILL_RADIUS)

    pil_img = Image.fromarray(img)

    subj_boxes = build_subjective_boxes(subj_starts)

    draw_text_on_image(pil_img, ANSWERS["subjective"], subj_boxes, FONT_PATH, FONT_SIZE, LINE_SPACING)

    temp_png = "temp_filled.png"

    pil_img.save(temp_png, dpi=(400, 400))

    with open(PDF_OUTPUT, "wb") as f:

        f.write(img2pdf.convert(temp_png, dpi=400))

    os.remove(temp_png)

    print(f"[SUCCESS] 输出文件: {PDF_OUTPUT}")

if __name__ == "__main__":

    main()