本文共 9426 字，大约阅读时间需要 31 分钟。

凸包

逼近多边形是轮廊的高度相似，但是有时候，我们希望使用一个多边形的凸包来简化它，凸包跟逼近多边形很相似，只不过它是物体最外层的“凸”多边形。凸包指的是完全包含原有轮廊，并且仅有轮廊上的点所构成的多边形。凸包的每一处都是凸的，即在凸包内连接任意两点的直线都在凸包的内部。在凸包内，任意连接三个点的内角小于180°

1.使用cv2.convexHull()函数，实现图像轮廓凸包的获取。 2.使用cv2.polylines()函数，实现图像轮廓凸包的绘制。

凸包的绘制

cv2.polylines(src, [points], isClosed,color, thickness)

src：要处理的图像。

points：点集。 isClosed：布尔型，True表示的是线段闭合，False表示的是仅保留线段。 color：线段颜色，格式是（B,G,R）值。 thickness：数值型，厚度，默认值为1，如果对封闭图形，正方形，三角形等传入-1,则会填充整个图形。

import cv2img = cv2.imread('contours2.png', 1)gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)ret, binary = cv2.threshold(gray, 127, 255, cv2.THRESH_BINARY)contours, hierarchy = cv2.findContours(binary,cv2.RETR_LIST,cv2.CHAIN_APPROX_NONE)cnt = contours[0]hull = cv2.convexHull(cnt)cv2.polylines(img,[hull],True,(255,0,0),2)cv2.imshow('line', img)cv2.waitKey()cv2.destroyAllWindows()

实物凸包检测

import cv2img = cv2.imread('hand.png', 1)gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)ret, binary = cv2.threshold(gray, 60, 255, cv2.THRESH_BINARY)contours, hierarchy = cv2.findContours(binary,cv2.RETR_LIST,cv2.CHAIN_APPROX_NONE)n=len(contours)    contoursImg=[]x=0 for i in range(n):    area = cv2.contourArea(contours[i])    if area>10000:            print(f"轮廓{i}的面积：\n{area}")        x=icnt = contours[x]cv2.imshow("binary",binary)hull = cv2.convexHull(cnt)cv2.polylines(img,[hull],True,(0,255,0),2)cv2.drawContours(img,contours,x,(0,0,255), 3)#显示图片cv2.imshow('line', img)cv2.waitKey(0)cv2.destroyAllWindows()

凸缺陷的概念

凸包与轮廓之间的部分，称为凸缺陷。轮廓上距离这条线最远的

点就是凸缺陷的点。

convexityDefects = cv2.convexityDefects(contour, convexhull)

contour：轮廓。

convexhull：凸包.

import cv2o = cv2.imread('contours2.png') gray = cv2.cvtColor(o,cv2.COLOR_BGR2GRAY) ret, binary = cv2.threshold(gray,127,255,cv2.THRESH_BINARY) contours, hierarchy = cv2.findContours(binary,cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)cnt = contours[0]hull = cv2.convexHull(cnt,returnPoints = False)            defects = cv2.convexityDefects(cnt,hull)         for i in range(defects.shape[0]):    s,e,f,d = defects[i,0]         start = tuple(cnt[s][0])           # 得到的是索引，要再轮廓中选出来    end = tuple(cnt[e][0])    far = tuple(cnt[f][0])    cv2.line(o,start,end,[0,0,255],2)    cv2.circle(o,far,5,[255,0,0],-1)cv2.imshow('result',o)cv2.waitKey(0)cv2.destroyAllWindows()

实物凸缺陷检测

import cv2import numpy as npimg = cv2.imread('hand.png') blur=cv2.medianBlur(img,5)gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY) ret, binary = cv2.threshold(gray,60,255,cv2.THRESH_BINARY) contours, hierarchy = cv2.findContours(binary, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)n=len(contours)       contoursImg=[]x=0 for i in range(n):    area = cv2.contourArea(contours[i])    if area>10000:            print(f"轮廓{i}的面积：\n{area}")        x=icnt = contours[x]hull = cv2.convexHull(cnt,returnPoints = False)            defects = cv2.convexityDefects(cnt,hull)         #print(defects)for i in range(defects.shape[0]):    s,e,f,d = defects[i,0]      #print(s,e,f,d)    start = tuple(cnt[s][0])              end = tuple(cnt[e][0])    far = tuple(cnt[f][0])    #print(far ,end)    cv2.line(img,start,end,[0,0,255],2)    cv2.circle(img,far,5,[255,0,0],-1)cv2.imshow('result',img)cv2.waitKey(0)cv2.destroyAllWindows()

宽高比

可以使用宽高比（AspectRation）来描述轮廓，例如矩形轮廓的宽高比为：

宽高比=宽度（Width）/高度（Height）

import cv2o = cv2.imread('binaryhand.png')  cv2.imshow("original",o)gray = cv2.cvtColor(o,cv2.COLOR_BGR2GRAY)  ret, binary = cv2.threshold(gray,127,255,cv2.THRESH_BINARY)  contours, hierarchy = cv2.findContours(binary,cv2.RETR_LIST,cv2.CHAIN_APPROX_SIMPLE)  x,y,w,h = cv2.boundingRect(contours[0])cv2.rectangle(o,(x,y),(x+w,y+h),(255,255,255),3)aspectRatio = float(w)/hprint(aspectRatio)cv2.imshow("result",o)cv2.waitKey()cv2.destroyAllWindows()

Extent

可以使用轮廓面积与矩形边界（矩形包围框、矩形轮廓）面积之比Extend来描述图像及其轮廓特征。计算方法为：

Extend =轮廓面积/矩形边界面积

import cv2o = cv2.imread('binaryhand.png')  cv2.imshow("original",o)gray = cv2.cvtColor(o,cv2.COLOR_BGR2GRAY)  ret, binary = cv2.threshold(gray,127,255,cv2.THRESH_BINARY)  contours, hierarchy = cv2.findContours(binary,cv2.RETR_LIST,cv2.CHAIN_APPROX_SIMPLE)  x,y,w,h = cv2.boundingRect(contours[0])cv2.drawContours(o,contours[0],-1,(0,0,255),3) cv2.rectangle(o,(x,y),(x+w,y+h),(255,0,0),3)rectArea=w*hcntArea=cv2.contourArea(contours[0])extend=float(cntArea)/rectAreaprint(extend)cv2.imshow("result",o)cv2.waitKey()cv2.destroyAllWindows()

最大值和最小值及它们的位置

OpenCV提供了函数cv2.minMaxLoc()，用于在指定的对象内查找最大值、最小值及其位置。该函数的语法格式是：

min_val, max_val, min_loc, max_loc = cv2.minMaxLoc(imgray, mask = mask)

min_val：最小值。

max_val：最大值。 min_loc：最小值的位置。 max_loc：最大值的位置。 OpenCV提供了函数cv2.mean()，用于计算一个对象的平均颜色或 平均灰度。该函数的语法格式为： mean_val = cv2.mean(im, mask = mask) im：原图像。 mask：掩模。

import cv2import numpy as npo = cv2.imread('ct2.jpg')  cv2.imshow("original",o)gray = cv2.cvtColor(o,cv2.COLOR_BGR2GRAY)  ret, binary = cv2.threshold(gray,127,255,cv2.THRESH_BINARY)  contours, hierarchy = cv2.findContours(binary,                                             cv2.RETR_LIST,                                             cv2.CHAIN_APPROX_SIMPLE)  cnt=contours[3]  mask = np.zeros(gray.shape,np.uint8)mask=cv2.drawContours(mask,[cnt],-1,255,-1)   minVal, maxVal, minLoc, maxLoc = cv2.minMaxLoc(gray,mask = mask)print("minVal=",minVal)print("maxVal=",maxVal)print("minLoc=",minLoc)print("maxLoc=",maxLoc)masko = np.zeros(o.shape,np.uint8)masko=cv2.drawContours(masko,[cnt],-1,(255,255,255),-1)loc=cv2.bitwise_and(o,masko) cv2.imshow("mask",loc)#显示灰度结果#loc=cv2.bitwise_and(gray,mask) #cv2.imshow("mask",loc)cv2.destroyAllWindows()

平均颜色及平均灰度

meanVal = cv2.mean(o,mask = mask)  #mask是区域，所以必须是单通道的print("meanVal=\n",meanVal)

极点

有时，我们希望获取某个对象内的极值点，例如最左端、最右端、

最上端、最下端的四个点。OpenCV提供了相应的函数来找出这些 点，通常的语法格式是：

leftmost = tuple(cnt[cnt[:, :,0].argmin()][0])    rightmost = tuple(cnt[cnt[:, :,0].argmax()][0])    topmost = tuple(cnt[cnt[:, :,1].argmin()][0])    bottommost = tuple(cnt[cnt[:, :,1].argmax()][0])

import cv2import numpy as npo = cv2.imread('binaryhand.png')  o1 = cv2.imread("face.png")o1 = cv2.resize(o1,(40,40))gray = cv2.cvtColor(o,cv2.COLOR_BGR2GRAY)  ret, binary = cv2.threshold(gray,127,255,cv2.THRESH_BINARY)  contours, hierarchy = cv2.findContours(binary,cv2.RETR_LIST,cv2.CHAIN_APPROX_SIMPLE)  mask = np.zeros(gray.shape,np.uint8)cnt=contours[0] cv2.drawContours(mask,[cnt],0,255,-1)leftmost = tuple(cnt[cnt[:,:,0].argmin()][0])rightmost = tuple(cnt[cnt[:,:,0].argmax()][0])topmost = tuple(cnt[cnt[:,:,1].argmin()][0])bottommost = tuple(cnt[cnt[:,:,1].argmax()][0])print("leftmost=",leftmost)print("rightmost=",rightmost)print("topmost=",topmost)print("bottommost=",bottommost)font=cv2.FONT_HERSHEY_SIMPLEXcv2.putText(o,'A',leftmost, font, 1,(0,0,255),2)o[284:324,229:269]=o1cv2.putText(o,'B',rightmost, font, 1,(0,0,255),2)cv2.putText(o,'C',topmost, font, 1,(0,0,255),2)cv2.putText(o,'D',bottommost, font, 1,(0,0,255),2)cv2.imshow("result",o)cv2.waitKey()cv2.destroyAllWindows()

轮廓特征值的应用场景

import cv2import numpy as npimg = cv2.imread('face.png')  cv2.imshow("original",img)gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)  ret, binary = cv2.threshold(gray,0,255,cv2.THRESH_BINARY)  contours, hierarchy = cv2.findContours(binary,cv2.RETR_LIST,cv2.CHAIN_APPROX_SIMPLE)  h,w,c=img.shapemask = np.zeros((h,w,c),np.uint8)for cnt in contours:        x,y,w,h = cv2.boundingRect(cnt)    cv2.rectangle(img,(x,y),(x+w,y+h),(255,255,255),3)    ratio = float(w)/h    if ratio<1:        cv2.drawContours(mask,[cnt],-1,(0,0,255), -1)        #n+=1        print(ratio,"是长的椭圆")    else:        cv2.drawContours(mask,[cnt],-1,(0,255,255), -1)        #n+=1        print(ratio,"是扁的椭圆")cv2.imshow("result",mask)cv2.waitKey()cv2.destroyAllWindows()

import cv2import numpy as npimg = cv2.imread('face.png')  cv2.imshow("original",img)gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)  ret, binary = cv2.threshold(gray,0,255,cv2.THRESH_BINARY)  contours, hierarchy = cv2.findContours(binary,cv2.RETR_LIST,cv2.CHAIN_APPROX_SIMPLE)  h,w,c=img.shapemask = np.zeros((h,w,c),np.uint8)for cnt in contours:        x,y,w,h = cv2.boundingRect(cnt)    cv2.rectangle(img,(x,y),(x+w,y+h),(255,255,255),3)    ratio = float(w)/h    rectArea=w*h    cntArea=cv2.contourArea(contours[0])    extend=float(cntArea)/rectArea    print(extend)        if ratio<1:        cv2.drawContours(mask,[cnt],-1,(0,0,255), -1)        print(ratio,"是长的椭圆")    else:        cv2.drawContours(mask,[cnt],-1,(0,255,255), -1)        print(ratio,"是扁的椭圆")    if extend==0.9818696450428397:        cv2.drawContours(mask,[cnt],-1,(255,0,255), -1)        cv2.imshow("result",mask)cv2.waitKey()cv2.destroyAllWindows()

计数

import cv2import numpy as npimg = cv2.imread('face.png')  cv2.imshow("original",img)gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)  ret, binary = cv2.threshold(gray,0,255,cv2.THRESH_BINARY)  contours, hierarchy = cv2.findContours(binary,cv2.RETR_LIST,cv2.CHAIN_APPROX_SIMPLE)  h,w,c=img.shapemask = np.zeros((h,w,c),np.uint8)n1,n2,n3=0,0,0for cnt in contours:        x,y,w,h = cv2.boundingRect(cnt)    cv2.rectangle(img,(x,y),(x+w,y+h),(255,255,255),3)    ratio = float(w)/h    rectArea=w*h    cntArea=cv2.contourArea(contours[0])    extend=float(cntArea)/rectArea    print(extend)        if extend==0.9818696450428397:        n1+=1        cv2.drawContours(mask,[cnt],-1,(255,0,0), -1)    elif ratio<1 and extend!=0.9818696450428397:        cv2.drawContours(mask,[cnt],-1,(0,0,255), -1)        n2+=1    else:        cv2.drawContours(mask,[cnt],-1,(0,255,255), -1)        n3+=1print(f"有{n1}个矩形，有{n2}个长的椭圆，有{n3}个扁的椭圆")cv2.imshow("result",mask)cv2.waitKey()cv2.destroyAllWindows()

转载地址：http://clklf.baihongyu.com/