OpenCV 轮廓检测

import cv2
import numpy as np
img0 = cv2.imread("cat.jpg")
#img = cv2.pyrUp(img)#面积放大4倍
img0 = cv2.pyrDown(img0)#原图有点大，面积缩小到1/4

img =255- img0  #黑白颠倒，视情况（白色（亮）背景则需要颠倒，使背景变黑色

gray = cv2.cvtColor(img.copy(), cv2.COLOR_BGR2GRAY) #转灰度图。防止后续更改了原图，函数参数使用原图的拷贝

#threshold(src, thresh, maxval, type[, dst]) -> retval, dst
ret, thresh = cv2.threshold( gray, 100, 255,cv2.THRESH_BINARY) #转二值图,亮度小于100的变为0，打于的变为255
cv2.imshow("binary", thresh)

#2个返回值，分别是轮廓的点集(contours)和各层轮廓的索引(hierarchy) # openCV 4 , 否则注意版本差异！
contours, hierarchy = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

for c in contours: #对于每一个轮廓
    #无倾斜的 边界矩形框
    x, y, w, h = cv2.boundingRect(c)
    cv2.rectangle(img0, (x,y), (x+w, y+h), (255,0,0), 2) # 颜色均为（B,G,R)模式

cv2.imshow("contours", img0)

W,H = img0.shape[0], img0.shape[1]
for c in contours: #对于每一个轮廓
    #无倾斜的 边界矩形框
    x, y, w, h = cv2.boundingRect(c)
    if w>0.1 *W and h >0.1*H:
        cv2.rectangle(img0, (x,y), (x+w, y+h), (255,0,0), 2) # 颜色均为（B,G,R)模式

W,H = img0.shape[0], img0.shape[1]
for c in contours: #对于每一个轮廓    # 可倾斜的最小边界矩形框
    rect = cv2.minAreaRect(c)
    box = cv2.boxPoints(rect)#计算出来的顶点，浮点型
    box = np.int0(box) #4个顶点坐标转化为整数
     #过滤太小的
    d =  np.linalg.norm(box[0]-box[2]) #对角线长度
    if d> 0.075* np.sqrt(W**2+H**2):
        cv2.drawContours(img0, [box], 0, (0,0,255),2)
cv2.imshow("contours", img0)

W,H = img0.shape[0], img0.shape[1]
for c in contours: #对于每一个轮廓    #最小边界圆
    (x,y) , radius = cv2.minEnclosingCircle(c)
    center, radius  = (int(x), int(y)) , int(radius)
    if radius > 0.05* np.sqrt(W**2+H**2):
        cv2.circle(img0, center, radius, (255,255,0),3)
cv2.imshow("contours", img0)

for c in contours: #对于每一个轮廓    
    epsilon = 0.001*cv2.arcLength(c, True) #epsilon为近似轮廓多边形周长与原轮廓周长的最大差值
    approxPloy = cv2.approxPolyDP(c, epsilon, True) #True表示 多边形封闭
    cv2.polylines(img0, [approxPloy], True, (255, 0, 255), 2) #
cv2.imshow("contours", img0)

    #凸轮廓需要先加上下面这句
    hull = cv2.convexHull(c) # 计算凸轮廓多变形时需要先处理轮廓