当前位置:   article > 正文

项目实战——答题卡识别判卷代码个人学习笔记

答题卡识别判卷

一、实现逻辑

1. 导入库并设置参数和正确答案;

2. 图像预处理:灰度=>滤波去噪=>Canny边缘算子处理;

3. 找到并扶正目标区域(试卷):检测并绘制所有轮廓=>筛选出试卷轮廓=>透视变换;

4. 在透视变换后的图中寻找试卷中的选项轮廓:二值化=>寻找轮廓=>筛选轮廓=>轮廓排序;

5. 答案匹配:在循环结构中,根据每个选项圆圈区域的像素占比判断所选的答案,并与正确

                      答案进行比对,最后格式化输出。

二、代码实现

 此项目中,环境依然使用的是信用卡实战项目中使用的pytorch环境。

所有自定义的函数都是前面信用卡实战OCR识别中用过的,所以不再进行解释,整体代码也无过多难以理解的地方,具体见如下详细注释

  1. #导入工具包
  2. import numpy as np
  3. import argparse
  4. import imutils
  5. import cv2
  6. # 设置参数
  7. ap = argparse.ArgumentParser()
  8. ap.add_argument("-i", "--image", required=True,
  9. help="path to the input image")
  10. args = vars(ap.parse_args())
  11. # 正确答案
  12. ANSWER_KEY = {0: 1, 1: 4, 2: 0, 3: 3, 4: 1}
  13. def order_points(pts):
  14. # 一共4个坐标点
  15. rect = np.zeros((4, 2), dtype = "float32")
  16. # 按顺序找到对应坐标0123分别是 左上,右上,右下,左下
  17. # 计算左上,右下
  18. s = pts.sum(axis = 1)
  19. rect[0] = pts[np.argmin(s)]
  20. rect[2] = pts[np.argmax(s)]
  21. # 计算右上和左下
  22. diff = np.diff(pts, axis = 1)
  23. rect[1] = pts[np.argmin(diff)]
  24. rect[3] = pts[np.argmax(diff)]
  25. return rect
  26. def four_point_transform(image, pts):
  27. # 获取输入坐标点
  28. rect = order_points(pts)
  29. (tl, tr, br, bl) = rect
  30. # 计算输入的w和h值
  31. widthA = np.sqrt(((br[0] - bl[0]) ** 2) + ((br[1] - bl[1]) ** 2))
  32. widthB = np.sqrt(((tr[0] - tl[0]) ** 2) + ((tr[1] - tl[1]) ** 2))
  33. maxWidth = max(int(widthA), int(widthB))
  34. heightA = np.sqrt(((tr[0] - br[0]) ** 2) + ((tr[1] - br[1]) ** 2))
  35. heightB = np.sqrt(((tl[0] - bl[0]) ** 2) + ((tl[1] - bl[1]) ** 2))
  36. maxHeight = max(int(heightA), int(heightB))
  37. # 变换后对应坐标位置
  38. dst = np.array([
  39. [0, 0],
  40. [maxWidth - 1, 0],
  41. [maxWidth - 1, maxHeight - 1],
  42. [0, maxHeight - 1]], dtype = "float32")
  43. # 计算变换矩阵
  44. M = cv2.getPerspectiveTransform(rect, dst)
  45. warped = cv2.warpPerspective(image, M, (maxWidth, maxHeight))
  46. # 返回变换后结果
  47. return warped
  48. def sort_contours(cnts, method="left-to-right"):
  49. reverse = False
  50. i = 0
  51. if method == "right-to-left" or method == "bottom-to-top":
  52. reverse = True
  53. if method == "top-to-bottom" or method == "bottom-to-top":
  54. i = 1
  55. boundingBoxes = [cv2.boundingRect(c) for c in cnts]
  56. (cnts, boundingBoxes) = zip(*sorted(zip(cnts, boundingBoxes),
  57. key=lambda b: b[1][i], reverse=reverse))
  58. return cnts, boundingBoxes
  59. def cv_show(name,img):
  60. cv2.imshow(name, img)
  61. cv2.waitKey(0)
  62. cv2.destroyAllWindows()
  63. # 预处理
  64. image = cv2.imread(args["image"])
  65. contours_img = image.copy()
  66. #灰度
  67. gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
  68. #滤波
  69. blurred = cv2.GaussianBlur(gray, (5, 5), 0)
  70. cv_show('blurred',blurred)
  71. #计算边缘
  72. edged = cv2.Canny(blurred, 75, 200)
  73. cv_show('edged',edged)
  74. # 轮廓检测
  75. cnts = cv2.findContours(edged.copy(), cv2.RETR_EXTERNAL,
  76. cv2.CHAIN_APPROX_SIMPLE)[1]
  77. cv2.drawContours(contours_img,cnts,-1,(0,0,255),3) #红色线框绘出轮廓
  78. cv_show('contours_img',contours_img)
  79. docCnt = None
  80. # 确保检测到了
  81. if len(cnts) > 0:
  82. # 根据轮廓面积大小进行排序
  83. cnts = sorted(cnts, key=cv2.contourArea, reverse=True)
  84. # 遍历每一个轮廓
  85. for c in cnts:
  86. # 近似
  87. peri = cv2.arcLength(c, True)
  88. approx = cv2.approxPolyDP(c, 0.02 * peri, True)
  89. # 准备做透视变换
  90. if len(approx) == 4:
  91. docCnt = approx
  92. break
  93. # 执行透视变换
  94. warped = four_point_transform(gray, docCnt.reshape(4, 2))
  95. cv_show('warped',warped)
  96. # Otsu's 阈值处理-二值化
  97. thresh = cv2.threshold(warped, 0, 255,
  98. cv2.THRESH_BINARY_INV | cv2.THRESH_OTSU)[1]
  99. cv_show('thresh',thresh)
  100. thresh_Contours = thresh.copy()
  101. # 找到每一个圆圈轮廓
  102. cnts = cv2.findContours(thresh.copy(), cv2.RETR_EXTERNAL,
  103. cv2.CHAIN_APPROX_SIMPLE)[1]
  104. cv2.drawContours(thresh_Contours,cnts,-1,(0,0,255),3) #因为是单通道,所以不显示轮廓
  105. cv_show('thresh_Contours',thresh_Contours)
  106. questionCnts = []
  107. # 遍历
  108. for c in cnts:
  109. # 计算比例和大小
  110. (x, y, w, h) = cv2.boundingRect(c)
  111. ar = w / float(h)
  112. # 根据实际情况指定标准
  113. if w >= 20 and h >= 20 and ar >= 0.9 and ar <= 1.1:
  114. questionCnts.append(c)
  115. # 按照从上到下进行排序
  116. questionCnts = sort_contours(questionCnts,
  117. method="top-to-bottom")[0]
  118. correct = 0
  119. # 每排有5个选项
  120. #q为enumerate对象中的索引;i为对应索引的真实值
  121. for (q, i) in enumerate(np.arange(0, len(questionCnts), 5)):
  122. # 排序
  123. cnts = sort_contours(questionCnts[i:i + 5])[0]
  124. bubbled = None
  125. # 遍历每一个结果
  126. for (j, c) in enumerate(cnts):
  127. # 使用mask来判断结果
  128. mask = np.zeros(thresh.shape, dtype="uint8") #用0数组建立掩膜mask
  129. #在mask上依次遍历每个选项的轮廓
  130. cv2.drawContours(mask, [c], -1, 255, -1) #-1表示填充
  131. cv_show('mask',mask)
  132. # 通过计算非零点数量来算是否选择这个答案
  133. #将thresh和mask进行图像的“与”运算,即只显示掩膜内的部分
  134. mask = cv2.bitwise_and(thresh, thresh, mask=mask)
  135. #计算像素值不为0的像素数
  136. total = cv2.countNonZero(mask)
  137. # 通过阈值判断每一行哪个选项被涂黑
  138. if bubbled is None or total > bubbled[0]:
  139. bubbled = (total, j)
  140. # 对比正确答案
  141. color = (0, 0, 255)
  142. k = ANSWER_KEY[q]
  143. # 判断是否正确
  144. if k == bubbled[1]:
  145. color = (0, 255, 0) #这一步貌似没什么必要
  146. correct += 1
  147. # 绘图
  148. cv2.drawContours(warped, [cnts[k]], -1, color, 3)
  149. score = (correct / 5.0) * 100
  150. print("[INFO] score: {:.2f}%".format(score))
  151. cv2.putText(warped, "{:.2f}%".format(score), (10, 30),
  152. cv2.FONT_HERSHEY_SIMPLEX, 0.9, (0, 0, 255), 2)
  153. cv2.imshow("Original", image)
  154. cv2.imshow("Exam", warped)
  155. cv2.waitKey(0)

此项目到处结束。

声明:本文内容由网友自发贡献,不代表【wpsshop博客】立场,版权归原作者所有,本站不承担相应法律责任。如您发现有侵权的内容,请联系我们。转载请注明出处:https://www.wpsshop.cn/w/人工智能uu/article/detail/768328
推荐阅读
相关标签
  

闽ICP备14008679号