python+opencv實現檢測物體聚集區域

內容涉及：二值圖像轉換 / 檢測連通區域面積 / 在原圖上畫框等

import cv2
import numpy as np

for n in open('list.txt'): # list.txt為目標文件列表
    path = n[:-1] # 去除文件路徑的換行符
    img = cv2.imread(path)
    gray =cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) # 圖像轉灰度
    ret, binary = cv2.threshold(gray, 75, 255, cv2.THRESH_BINARY) # 灰度轉二值圖像
    cv2.imwrite(path + 'abc.png', binary)
    kernel = np.ones((21,21),np.uint8) # 給圖像閉運算定義核
    kernel_1 = np.ones((101,101),np.uint8) # 給圖像開運算定義核
    # 圖像先閉運算再開運算可以過濾孤立的物體， 將密集物體區域形成一片連通區
    closing = cv2.morphologyEx(binary, cv2.MORPH_CLOSE, kernel)
    opening = cv2.morphologyEx(closing, cv2.MORPH_OPEN, kernel_1)
    # 給圖像的邊緣像素設定為255，否則下面連通區的檢測無法識別貼在圖像邊緣的連通區
    # 特別注意！！！，此操作會將整個圖像也視為一個連通區域
    opening_x = opening.shape[0]
    opening_y = opening.shape[1]
    opening[:,0] = 255
    opening[:,opening_y-1] = 255
    opening[0,:] = 255
    opening[opening_x-1,:] = 255
    # 檢測圖像連通區（輸入為二值化圖像）
    image, contours, hierarchy = cv2.findContours(opening,1,2)
    for n in range(len(contours)):
        # 篩選面積較大的連通區，閾值為20000
        cnt = contours[n]
        area = cv2.contourArea(cnt)
        if area > 20000:
            x,y,w,h=cv2.boundingRect(cnt)
            img_ = cv2.rectangle(img ,(x,y),(x+w,y+h),(0,0,255),4) # 畫框
            print('')
            img__ = img[y-h:y+h,x-w:x+w,:]
    cv2.imwrite(path + 'abc_open.png', opening)
    cv2.imwrite(path + 'abc_close.png', closing)
    cv2.imwrite(path + 'abc_close_range.png', img_)