【深度学习】OpenCV 实战：答题卡识别 项目流程详解

m沐沐2026-06-29 17:27

文章目录

完整代码一览

c 复制代码 
import numpy as np
import cv2

def cv_show(name, image):
    cv2.imshow(name, image)
    cv2.waitKey(0)

def resize(image, width=None, height=None, inter=cv2.INTER_AREA):
    dim = None
    (h, w) = image.shape[:2]
    if width is None and height is None:
        return image
    if width is None:
        r = height / float(h)
        dim = (int(w * r), height)
    else:
        r = width / float(w)
        dim = (width, int(h * r))
    resized = cv2.resize(image, dim, interpolation=inter)
    return resized

def order_points(pts):
    rect = np.zeros((4, 2), dtype='float32')
    s = pts.sum(axis=1)
    rect[0] = pts[np.argmin(s)]   # 左上
    rect[2] = pts[np.argmax(s)]   # 右下
    diff = np.diff(pts, axis=1)
    rect[1] = pts[np.argmin(diff)] # 右上
    rect[3] = pts[np.argmax(diff)] # 左下
    return rect

def four_point_transform(image, pts):
    rect = order_points(pts)
    (tl, tr, br, bl) = rect
    widthA = np.sqrt(((br[0] - bl[0]) ** 2) + ((br[1] - bl[1]) ** 2))
    widthB = np.sqrt(((tr[0] - tl[0]) ** 2) + ((tr[1] - tl[1]) ** 2))
    maxWidth = max(int(widthA), int(widthB))
    heightA = np.sqrt(((tr[0] - br[0]) ** 2) + ((tr[1] - br[1]) ** 2))
    heightB = np.sqrt(((tl[0] - bl[0]) ** 2) + ((tl[1] - bl[1]) ** 2))
    maxHeight = max(int(heightA), int(heightB))
    dst = np.array([[0, 0], [maxWidth - 1, 0], [maxWidth - 1, maxHeight - 1], [0, maxHeight - 1]], dtype='float32')
    M = cv2.getPerspectiveTransform(rect, dst)
    warped = cv2.warpPerspective(image, M, (maxWidth, maxHeight))
    return warped

def sort_contours(cnts, method='left-to-right'):
    reverse = False
    i = 0
    if method == 'right-to-left' or method == 'bottom-to-top':
        reverse = True
    if method == 'top-to-bottom' or method == 'bottom-to-top':
        i = 1
    boundingBoxes = [cv2.boundingRect(c) for c in cnts]
    (cnts, boundingBoxes) = zip(*sorted(zip(cnts, boundingBoxes),
                                        key=lambda b: b[1][i], reverse=reverse))
    return cnts, boundingBoxes

ANSWER_KEY = {0: 1, 1: 4, 2: 0, 3: 3, 4: 1}   # 5道题的正确答案（0-4对应A-E）

# ---------- 主程序 ----------
image = cv2.imread(r'./images/images/test_01.png')
contours_img = image.copy()
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
blurred = cv2.GaussianBlur(gray, ksize=(5, 5), sigmaX=0)
cv_show('blurred', blurred)
edged = cv2.Canny(blurred, threshold1=75, threshold2=200)
cv_show('edged', edged)

cnts = cv2.findContours(edged.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)[-2]
cv2.drawContours(contours_img, cnts, -1, (0, 0, 255), 3)
cv_show('contours_img', contours_img)

docCnt = None
cnts = sorted(cnts, key=cv2.contourArea, reverse=True)
for c in cnts:
    peri = cv2.arcLength(c, closed=True)
    approx = cv2.approxPolyDP(c, 0.02 * peri, closed=True)
    if len(approx) == 4:
        docCnt = approx
        break

warped_t = four_point_transform(image, docCnt.reshape(4, 2))
warped_new = warped_t.copy()
cv_show('warped', warped_t)
warped = cv2.cvtColor(warped_t, cv2.COLOR_BGR2GRAY)

thresh = cv2.threshold(warped, 0, 255, cv2.THRESH_BINARY_INV | cv2.THRESH_OTSU)[1]
cv_show('thresh', thresh)
thresh_Contours = thresh.copy()

cnts = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)[-2]
warped_Contours = cv2.drawContours(warped_t, cnts, -1, (0, 255, 0), 1)
cv_show('warped_Contours', warped_Contours)

questionCnts = []
for c in cnts:
    x, y, w, h = cv2.boundingRect(c)
    ar = w / float(h)
    if w >= 20 and h >= 20 and 0.9 <= ar <= 1.1:
        questionCnts.append(c)
print(len(questionCnts))

questionCnts = sort_contours(questionCnts, method="top-to-bottom")[0]
correct = 0

for q, i in enumerate(np.arange(0, len(questionCnts), 5)):
    cnts = sort_contours(questionCnts[i:i + 5])[0]
    bubbled = None
    for j, c in enumerate(cnts):
        mask = np.zeros(thresh.shape, dtype="uint8")
        cv2.drawContours(mask, [c], -1, 255, -1)
        cv_show('mask', mask)
        thresh_mask_and = cv2.bitwise_and(thresh, thresh, mask=mask)
        cv_show('thresh_mask_and', thresh_mask_and)
        total = cv2.countNonZero(thresh_mask_and)
        if bubbled is None or total > bubbled[0]:
            bubbled = (total, j)
    color = (0, 0, 255)
    k = ANSWER_KEY[q]
    if k == bubbled[1]:
        color = (0, 255, 0)
        correct += 1
    cv2.drawContours(warped_new, [cnts[k]], -1, color, 3)
    cv_show('warpping', warped_new)

score = (correct / 5.0) * 100
print("[INFO] score: {:.2f}%".format(score))
cv2.putText(warped_new, "{:.2f}%".format(score), (10, 30),
            cv2.FONT_HERSHEY_SIMPLEX, 0.9, (0, 0, 255), 2)
cv2.imshow("Original", image)
cv2.imshow("Exam", warped_new)
cv2.waitKey(0)

导入库与辅助函数

1.图像显示函数cv_show

c 复制代码 
import numpy as np
import cv2

def cv_show(name, image):
    cv2.imshow(name, image)
    cv2.waitKey(0)

2.图像缩放函数 resize

c 复制代码 
def resize(image, width=None, height=None, inter=cv2.INTER_AREA):
    dim = None
    (h, w) = image.shape[:2]
    if width is None and height is None:
        return image
    if width is None:
        r = height / float(h)
        dim = (int(w * r), height)
    else:
        r = width / float(w)
        dim = (width, int(h * r))
    resized = cv2.resize(image, dim, interpolation=inter)
    return resized

3. 坐标排序 order_points

c 复制代码 
def order_points(pts):
    rect = np.zeros((4, 2), dtype='float32')
    s = pts.sum(axis=1)          # x+y
    rect[0] = pts[np.argmin(s)]  # 左上角
    rect[2] = pts[np.argmax(s)]  # 右下角
    diff = np.diff(pts, axis=1)  # x-y
    rect[1] = pts[np.argmin(diff)] # 右上角
    rect[3] = pts[np.argmax(diff)] # 左下角
    return rect

4.透视变换 four_point_transform

将任意四边形区域通过透视变换拉正为矩形，用于矫正答题卡视角。

c 复制代码 
def four_point_transform(image, pts):
    rect = order_points(pts)
    (tl, tr, br, bl) = rect
    # 计算目标宽度（取上下边距离的最大值）
    widthA = np.sqrt(((br[0] - bl[0]) ** 2) + ((br[1] - bl[1]) ** 2))
    widthB = np.sqrt(((tr[0] - tl[0]) ** 2) + ((tr[1] - tl[1]) ** 2))
    maxWidth = max(int(widthA), int(widthB))
    # 计算目标高度（取左右边距离的最大值）
    heightA = np.sqrt(((tr[0] - br[0]) ** 2) + ((tr[1] - br[1]) ** 2))
    heightB = np.sqrt(((tl[0] - bl[0]) ** 2) + ((tl[1] - bl[1]) ** 2))
    maxHeight = max(int(heightA), int(heightB))
    # 目标矩形四个点（正视图）
    dst = np.array([[0, 0], [maxWidth - 1, 0], [maxWidth - 1, maxHeight - 1], [0, maxHeight - 1]], dtype='float32')
    M = cv2.getPerspectiveTransform(rect, dst)
    warped = cv2.warpPerspective(image, M, (maxWidth, maxHeight))
    return warped

4.轮廓排序 sort_contours

c 复制代码 
def sort_contours(cnts, method='left-to-right'):
    reverse = False
    i = 0
    if method == 'right-to-left' or method == 'bottom-to-top':
        reverse = True
    if method == 'top-to-bottom' or method == 'bottom-to-top':
        i = 1
    boundingBoxes = [cv2.boundingRect(c) for c in cnts]
    (cnts, boundingBoxes) = zip(*sorted(zip(cnts, boundingBoxes),
                                        key=lambda b: b[1][i], reverse=reverse))
    return cnts, boundingBoxes

5.标准答案

5 道题，每道题的正确选项索引（0=A, 1=B, 2=C, 3=D, 4=E）。

c 复制代码 
ANSWER_KEY = {0: 1, 1: 4, 2: 0, 3: 3, 4: 1}

主程序流程

图像预处理

c 复制代码 
image = cv2.imread(r'./images/images/test_01.png')
contours_img = image.copy()
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
blurred = cv2.GaussianBlur(gray, ksize=(5, 5), sigmaX=0)
cv_show('blurred', blurred)
edged = cv2.Canny(blurred, threshold1=75, threshold2=200)
cv_show('edged', edged)

先读取图像，转为灰度图，再高斯模糊（5×5 核）降低噪声，避免边缘检测受噪点干扰，最后Canny 边缘检测，阈值 75 和 200 为经验值，可调整。

运行结果：

轮廓检测并筛选最大外轮廓

c 复制代码 
cnts = cv2.findContours(edged.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)[-2]
cv2.drawContours(contours_img, cnts, -1, (0, 0, 255), 3)
cv_show('contours_img', contours_img)

docCnt = None
cnts = sorted(cnts, key=cv2.contourArea, reverse=True)
for c in cnts:
    peri = cv2.arcLength(c, closed=True)
    approx = cv2.approxPolyDP(c, 0.02 * peri, closed=True)
    if len(approx) == 4:
        docCnt = approx
        break

检测所有外层轮廓，绘制出来，同时按面积从大到小排序，遍历轮廓，用多边形近似，找到第一个近似为四边形（4 个顶点）的轮廓，即为答题卡外框。

运行结果：

透视矫正与二值化

c 复制代码 
warped_t = four_point_transform(image, docCnt.reshape(4, 2))
warped_new = warped_t.copy()
cv_show('warped', warped_t)
warped = cv2.cvtColor(warped_t, cv2.COLOR_BGR2GRAY)

thresh = cv2.threshold(warped, 0, 255, cv2.THRESH_BINARY_INV | cv2.THRESH_OTSU)[1]
cv_show('thresh', thresh)

将答题卡区域拉正为矩形，转为灰度图，然后使用 OTSU 自动阈值二值化，并反转（BINARY_INV），使选项涂黑区域变为白色（方便统计白色像素），背景为黑色。

运行结果：

提取所有圆形选项轮廓并排序

c 复制代码 
cnts = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)[-2]
warped_Contours = cv2.drawContours(warped_t, cnts, -1, (0, 255, 0), 1)
cv_show('warped_Contours', warped_Contours)

questionCnts = []
for c in cnts:
    x, y, w, h = cv2.boundingRect(c)
    ar = w / float(h)
    if w >= 20 and h >= 20 and 0.9 <= ar <= 1.1:
        questionCnts.append(c)
print(len(questionCnts))
questionCnts = sort_contours(questionCnts, method="top-to-bottom")[0]

在二值图上再次检测轮廓，得到所有白色区域（即被涂黑的选项）。过滤条件：外接矩形宽高不小于 20 像素，且长宽比接近 1:1（圆形选项），筛选出所有有效选项轮廓。按从上到下排序所有选项轮廓。

运行结果：

标记正误

c 复制代码 
correct = 0

for q, i in enumerate(np.arange(0, len(questionCnts), 5)):
    cnts = sort_contours(questionCnts[i:i + 5])[0]
    bubbled = None
    for j, c in enumerate(cnts):
        mask = np.zeros(thresh.shape, dtype="uint8")
        cv2.drawContours(mask, [c], -1, 255, -1)
        cv_show('mask', mask)
        thresh_mask_and = cv2.bitwise_and(thresh, thresh, mask=mask)
        cv_show('thresh_mask_and', thresh_mask_and)
        total = cv2.countNonZero(thresh_mask_and)
        if bubbled is None or total > bubbled[0]:
            bubbled = (total, j)
    color = (0, 0, 255)
    k = ANSWER_KEY[q]
    if k == bubbled[1]:
        color = (0, 255, 0)
        correct += 1
    cv2.drawContours(warped_new, [cnts[k]], -1, color, 3)
    cv_show('warpping', warped_new)

遍历每 5 个轮廓（即一道题），对这 5 个轮廓按从左到右排序，对每个选项创建掩码，与二值图做按位与，统计白色像素数（即涂黑面积）。

选出面积最大的选项作为考生的答案，记录其索引 j，与标准答案对比：若正确，用绿色框标记，否则红色框。

运行结果：

打印正确率并批注

c 复制代码 
score = (correct / 5.0) * 100
print("[INFO] score: {:.2f}%".format(score))
cv2.putText(warped_new, "{:.2f}%".format(score), (10, 30),
            cv2.FONT_HERSHEY_SIMPLEX, 0.9, (0, 0, 255), 2)
cv2.imshow("Exam", warped_new)
cv2.waitKey(0)

在矫正后的答题卡图像左上角打印分数。

运行结果：

热门推荐
012026年6月AI大模型全景报告：GPT-5.6、Claude Opus 4.8、Gemini 3.5，中美AI三足鼎立谁主沉浮？022026年6月AI行业全景：从百模大战到Agent元年，这30天发生了什么？032026 年 AI 编程工具终极横评：Cursor vs Claude Code vs Copilot vs Windsurf04【AI】2026 年具身智能模型和世界模型总结05Trae国际版与国内版深度测评：AI原生IDE的双生花06飞书长连接_事件订阅（接收消息，审批任务状态变更）07Claude Code、Codex、Cursor三分天下：2026年AI编程Agent生态全景剖析08GitHub 镜像站点092026 AI 编程工具终极实战指南：Cursor vs Claude Code vs Copilot，开发者该怎么选？102026年AI架构实战：彻底解决OpenAI接口超时与封号，Python调用GPT-5.2/Sora2企业级架构详解（附源码+压测报告）