套件安裝

1. scipy：pip install scipy
2. imutils：pip install imutils

練習用圖

1. edge.jpg

內容

1. OpenCV邊緣偵測

   1.1 影像前處理

   def preprocessing(image):
       # 灰階
       gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
       show_img('gray', gray)
   
       # 高斯濾波
       gaussian = cv2.GaussianBlur(gray, (9, 9), 0)
       show_img("gaussian", gaussian)
   
       # 開運算去除白色噪點
       kernel = np.ones((9, 9), np.uint8)
       open = cv2.morphologyEx(gaussian, cv2.MORPH_OPEN, 
                               kernel, iterations=3)
       show_img("open", open)
       return open
   

   1.2 Canny影像邊緣檢測

   def edge_detect(image):
       # 以canny邊緣檢測算法獲取目標(50~100為閾值)
       edged = cv2.Canny(image, 50, 100) #低於50刪除 高於100留下
       show_img("edged", edged)
   
       # 在邊緣圖像中尋找物體輪廓
       contours1, hierarchy = cv2.findContours(edged.copy(), 
                                               cv2.RETR_EXTERNAL,
                                               cv2.CHAIN_APPROX_SIMPLE)
   
       # 物體輪廓由左到右進行排序
       (contours2, _) = contours.sort_contours(contours1)
       # 若輪廓面積小於100，視為噪音濾除
       contours2 = [i for i in contours2 if cv2.contourArea(i) > 100]
   
       # 初始化pixelsPerMetric
       pixelsPerMetric = None
   
       return contours2, pixelsPerMetric
   

   1.3 計算輪廓長度高度，並畫出外切線框

   # 計算物體像素尺寸，並轉換為實際尺寸
   def measure(image, contours2, pixelsPerMetric):
       origin = image.copy()
       for i in contours2:
           # 計算出物品輪廓之外切線框
           box = cv2.minAreaRect(i)
           box = cv2.cv.BoxPoints(box) if imutils.is_cv2()  \
                 else cv2.boxPoints(box)
           box = np.array(box, dtype="int")
   
           # 左上角座標開始順時針排序，並畫出外切線框
           box = perspective.order_points(box)
           cv2.drawContours(origin, [box.astype("int")], -1, (0, 255, 0), 2)
   
           # 畫書外切線框端點
           for (x, y) in box:
               cv2.circle(origin, (int(x), int(y)), 5, (0, 0, 255), -1)
   
           #算出左上和右上端點的中心點、左下和右下端點的中心點
           (tl, tr, br, bl) = box
           (tltrX, tltrY) = midpoint(tl, tr)
           (blbrX, blbrY) = midpoint(bl, br)
   
           # 算出左上和左下端點的中心點、右上和右下端點的中心點
           (tlblX, tlblY) = midpoint(tl, bl)
           (trbrX, trbrY) = midpoint(tr, br)
   
           # 計算兩個中心點距離(dA：寬、dB：長)
           dA = dist.euclidean((tltrX, tltrY), (blbrX, blbrY))
           dB = dist.euclidean((tlblX, tlblY), (trbrX, trbrY))
   
           # 像素值與最左邊物品實際長度比值
           if pixelsPerMetric is None:
               pixelsPerMetric = dB / image_width
   
           # 計算目標的實際大小
           dimA = dA / pixelsPerMetric
           dimB = dB / pixelsPerMetric
   
           # 在圖片中標註物體尺寸
           cv2.putText(origin, "{:.1f}cm".format(dimB),
                       (int(tltrX - 15), int(tltrY - 10)), 
                       cv2.FONT_HERSHEY_SIMPLEX,
                       0.65, (255, 255, 255), 2)
           cv2.putText(origin, "{:.1f}cm".format(dimA),
                       (int(trbrX - 10), int(trbrY)),
                       cv2.FONT_HERSHEY_SIMPLEX,
                       0.65, (255, 255, 255), 2)
   
       show_img("Image", origin)
       return origin
   

   1.4 完整程式碼

   from scipy.spatial import distance as dist
   from imutils import perspective
   from imutils import contours
   import numpy as np
   import imutils
   import cv2
   
   # 讀取中文路徑圖檔(圖片讀取為BGR)
   def cv_imread(image_path):
       image = cv2.imdecode(np.fromfile(image_path, dtype=np.uint8), -1)
       image = cv2.cvtColor(image, cv2.COLOR_BGRA2BGR)
       return image
   
   # 顯示圖檔
   def show_img(name, image):
       cv2.imshow(name, image)
       cv2.waitKey(0)
   
   # 圖片預處理
   def preprocessing(image):
       # 灰階
       gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
       show_img('gray', gray)
   
       # 高斯濾波
       gaussian = cv2.GaussianBlur(gray, (9, 9), 0)
       show_img("gaussian", gaussian)
   
       # 開運算去除白色噪點
       kernel = np.ones((9, 9), np.uint8)
       open = cv2.morphologyEx(gaussian, cv2.MORPH_OPEN, 
                               kernel, iterations=3)
       show_img("open", open)
       return open
   
   # 計算物體中間點
   def midpoint(point1, point2):
       point = ((point1[0]+point2[0])/2, (point1[1]+point2[1])/2)
       return point
   
   def edge_detect(image):
       # 以canny邊緣檢測算法獲取目標(50~100為閾值)
       edged = cv2.Canny(image, 50, 100) #低於50刪除 高於100留下
       show_img("edged", edged)
   
       # 在邊緣圖像中尋找物體輪廓
       contours1, hierarchy = cv2.findContours(edged.copy(), 
                                               cv2.RETR_EXTERNAL,
                                               cv2.CHAIN_APPROX_SIMPLE)
       # 物體輪廓由左到右進行排序
       (contours2, _) = contours.sort_contours(contours1)
       # 若輪廓面積小於100，視為噪音濾除
       contours2 = [i for i in contours2 if cv2.contourArea(i) > 100]
   
       # 初始化 pixels per metric
       pixelsPerMetric = None
   
       return contours2, pixelsPerMetric
   
   # 計算物體像素尺寸，並轉換為實際尺寸
   def measure(image, contours2, pixelsPerMetric):
       origin = image.copy()
       for i in contours2:
           # 計算出物品輪廓之外切線框
           box = cv2.minAreaRect(i)
           box = cv2.cv.BoxPoints(box) if imutils.is_cv2() \
                 else cv2.boxPoints(box)
           box = np.array(box, dtype="int")
   
           # 左上角座標開始順時針排序，並畫出外切線框
           box = perspective.order_points(box)
           cv2.drawContours(origin, [box.astype("int")], -1, (0, 255, 0), 2)
   
           # 畫書外切線框端點
           for (x, y) in box:
               cv2.circle(origin, (int(x), int(y)), 5, (0, 0, 255), -1)
   
           # 算出左上和右上端點的中心點、左下和右下端點的中心點
           (tl, tr, br, bl) = box
           (tltrX, tltrY) = midpoint(tl, tr)
           (blbrX, blbrY) = midpoint(bl, br)
   
           # 算出左上和左下端點的中心點、右上和右下端點的中心點
           (tlblX, tlblY) = midpoint(tl, bl)
           (trbrX, trbrY) = midpoint(tr, br)
   
           # 計算兩個中心點距離(dA：寬、dB：長)
           dA = dist.euclidean((tltrX, tltrY), (blbrX, blbrY))
           dB = dist.euclidean((tlblX, tlblY), (trbrX, trbrY))
   
           # 像素值與最左邊物品實際長度比值
           if pixelsPerMetric is None:
               pixelsPerMetric = dB / image_width
   
           # 計算目標的實際大小
           dimA = dA / pixelsPerMetric
           dimB = dB / pixelsPerMetric
   
           # 在圖片中標註物體尺寸
           cv2.putText(origin, "{:.1f}cm".format(dimB),
                       (int(tltrX - 15), int(tltrY - 10)), 
                       cv2.FONT_HERSHEY_SIMPLEX,
                       0.65, (255, 255, 255), 2)
           cv2.putText(origin, "{:.1f}cm".format(dimA),
                       (int(trbrX - 10), int(trbrY)), 
                       cv2.FONT_HERSHEY_SIMPLEX,
                       0.65, (255, 255, 255), 2)
   
       show_img("Image", origin)
       return origin
   
   if __name__ == "__main__":
       image_path = './edge.jpg'
       image_width = 10.8
       image = cv_imread(image_path)
       show_img("image", image)
       gray = preprocessing(image)
       contours2, pixelsPerMetric = edge_detect(gray)
       measure(image, contours2, pixelsPerMetric)   
   

   1.5 執行結果

   • 讀取edge.jpg原圖
     

   • 將圖片轉換成灰階
     

   • 高斯模糊保留圖像主要輪廓，並降低躁點影響。
     

   • 開運算去除白色噪點(如：滑鼠上白色英文單字)
     

   • 以Canny邊緣檢測算法獲取待量測物體
     

   • 計算輪廓長度高度，並畫出外切線框
     

小結

1. 下一站，我們前往「OpenCV物件偵測」。

讓我們繼續看下去...

參考資料

1. Measuring size of objects in an image with OpenCV