opencv python 各种轮廓排序

OpenCV轮廓排序

contours.sort(key=cnt_area, reverse=True) #排序

reverse：True表示降序 False表示降序

 如果轮廓线是闭合的 就使用面积来排序使用

cv2.contourArea()  获得轮廓面积后进行排序：如下

import cv2
import numpy as np
from numpy import random as nr
from skimage import morphology,data,color
from skimage import img_as_float,img_as_ubyte

def imgContoursROI():
    """使用查找到的轮廓生成ROI剪切图片中感兴趣区域"""
    img = cv2.imread('JP1.JPG')  # 载入图像
    cv2.imshow("src", img)
    copyImg=img.copy()  #原图像的拷贝
    gray=cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    ret, binary = cv2.threshold(gray, 160, 255, cv2.THRESH_BINARY)  # 阈值化
    contours, hierarchy = cv2.findContours(binary, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)  # 寻找轮廓线
    contours.sort(key=cnt_area, reverse=True) #排序
    mask=np.zeros(img.shape, np.uint8) #原图大小的纯黑mask图像
    draw=np.zeros(img.shape, np.uint8) #将原图 copyTo 到draw上,加上mask操作
    temp=np.zeros(img.shape, np.uint8) #每次循环，重置mask 和 draw
    i=0
    for contour in contours: # 遍历所有轮廓
        area = cv2.contourArea(contour) #轮廓图面积
        B = 255
        G = 255
        R = 255 #生成颜色
        mask=temp.copy()
        draw=temp.copy()
        if (area > 300) : #轮廓面积大于300的做 mask
            x, y, w, h = cv2.boundingRect(contour)
            cv2.drawContours(img,[contour],-1,(B,G,R),cv2.FILLED)
            cv2.rectangle(img, (x, y), (x + w, y + h), (255, 0, 0), 2, cv2.LINE_AA)
            cv2.putText(img, "No.%d" % (i + 1), (x, y - 5), font, 0.8, (255, 0, 0), 2)
            #ROIname='ROI{num}'.format(num=i)
            #ROI = cv2.bitwise_and(img, mask) #图形与mask 的 and 运算
            cv2.imshow("dstimg", img)
            i+=1
    cv2.waitKey(0)


def cnt_area(cnt):
    """返回轮廓的面积"""
    area = cv2.contourArea(cnt)
    return area

效果图 

如果 轮廓不闭合，使用 长度来排序

 cv2.arcLength(cnt,False)

代码如下：

def circle_fitness():
    """拟合图像中的圆，然后 去掉失真的图像"""
    img = cv2.imread('JP1.JPG')  # 载入图像
    #cv2.imshow("src", img)
    copyImg=img.copy()  #原图像的拷贝
    mask = np.zeros(img.shape, np.uint8)  # 原图大小的纯黑mask图像
    mask1 = np.zeros(img.shape, np.uint8)  # 原图大小的纯黑mask图像

    maskw = np.zeros(img.shape, np.uint8)  # 原图大小的纯黑mask图像
    maskw[:]=255                            #生成白色mask图像
    gray=cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    cv2.namedWindow('src',cv2.WINDOW_AUTOSIZE)
    himg = cv2.equalizeHist(gray)  # 直方图均匀化
    while (True):
        ret, binary = cv2.threshold(himg, 132, 255, cv2.THRESH_BINARY)  # 阈值化
        contours, hierarchy = cv2.findContours(binary, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)  # 寻找轮廓线
        for contour in contours:  # 遍历所有轮廓
            area = cv2.contourArea(contour)  # 轮廓图面积
            B = nr.randint(1, 255)
            G = nr.randint(1, 255)
            R = nr.randint(1, 255)
            if (area > 200):  # 轮廓面积大于200的
                #ellipse = cv2.fitEllipse(contour) #根据轮廓拟合椭圆
                #cv2.ellipse(copyImg, ellipse, (B,G,R), 2, cv2.LINE_AA)
                #x, y = ellipse[0]
                circle=cv2.minEnclosingCircle(contour) #通过轮廓拟合最小外接圆
                Cx,Cy=circle[0]                        #获得外接圆圆心点
                r=circle[1]                            #获得外接圆半径
                cv2.circle(mask, (np.int(Cx), np.int(Cy)) ,np.int(r-10) ,(255, 255, 255), -1, 8, 0)
                ROI = cv2.bitwise_and(copyImg, mask)  # 图形与mask 的 and 运算
                #cv2.circle(copyImg, (np.int(x1), np.int(y1)), 4, (255, 0, 0), -1, 8, 0)

        grayROI = cv2.cvtColor(ROI, cv2.COLOR_BGR2GRAY)
        ret, binary = cv2.threshold(grayROI, 100, 150, cv2.THRESH_BINARY)  # 阈值化
        #binary = cv2.adaptiveThreshold(grayROI, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 63, 8)
        contours, hierarchy = cv2.findContours(binary, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)  # 寻找轮廓线
        cv2.drawContours(mask1, contours, -1, (255, 255, 255), cv2.FILLED)
        mask2=~mask1

        dst = cv2.bitwise_and(mask2, mask)  # 图形与mask 的 and 运算
        cv2.imshow("dst", dst)
        cv2.imshow("src", img)
        break
    return dst


def get_SkeletonSkimage(img):
    mask1 = np.zeros(img.shape, np.uint8)  # 原图大小的纯黑mask图像
    skim=opencv2skimage(img)           #转换图像格式opencv图像格式转skimage图像格式
    image = color.rgb2gray(skim)       #skimage转灰度图
    skeleton = morphology.skeletonize(image)
    dstimg=skimage2opencv(skeleton)       #转换图像格式skimage图像格式专opencv格式
    #gray=cv2.cvtColor(dstimg,cv2.COLOR_BGR2GRAY)          #opencv 转灰度图
    contours, hierarchy = cv2.findContours(dstimg, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)  # 寻找轮廓线
    contours.sort(key=cnt_length, reverse=True)
    #cv2.drawContours(mask1, contours, -1, (255, 255, 255), cv2.FILLED)
    #arrimgContours=np.array(contours)
    i=0
    for contour in contours:  # 遍历所有轮廓
        #arrimg=np.array(contour)
        #size=arrimg.size
        x, y, w, h = cv2.boundingRect(contour)
        #area = cv2.contourArea(contour)  # 轮廓图面积
        B = nr.randint(1, 255)
        G = nr.randint(1, 255)
        R = nr.randint(1, 255)
        cv2.rectangle(mask1, (x, y), (x + w, y + h), (255, 0, 0), 2, cv2.LINE_AA)
        cv2.putText(mask1, "No.%d" % (i + 1), (x, y - 5), font, 0.8, (255, 0, 0), 2)
        cv2.drawContours(mask1, [contour], -1, (B, G, R), cv2.FILLED)
        i += 1
    cv2.imshow('SkeletonContours', mask1)
    cv2.waitKey()


def cnt_length(cnt):
    """length"""
    length = cv2.arcLength(cnt,False)
    return length


if __name__ == "__main__":
    #myFindContours()
    #imgContoursROI()
    #getContoursInROI()
    #TrackbarThreshold()
    #get_SkeletonSkimage(circle_fitness())

效果图：