如何使用Python在OpenCV中找到轮廓的颜色

Question

我正在处理一个需求，需要找到轮廓内区域的颜色。我们使用 Python 和 OpenCv，以下是我的代码：

import imutils
import cv2
import numpy as np

path = "multiple_grains_1.jpeg"
img = cv2.imread(path)
resized = imutils.resize(img, width=900)
ratio = img.shape[0] / float(resized.shape[0])
gray = cv2.cvtColor(resized, cv2.COLOR_BGR2GRAY)

(ret, thresh) = cv2.threshold(gray, 0, 255, cv2.THRESH_BINARY | cv2.THRESH_OTSU)
edge = cv2.Canny(thresh, 100, 200)
( _,cnts, _) = cv2.findContours(edge.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

for c in cnts:
    rect = cv2.minAreaRect(c)
    box = cv2.boxPoints(rect)
    box = np.int0(box)
    area = cv2.contourArea(c)
    if area > 1:
        cv2.drawContours(resized,[box],0,(0,0,255),2)
        cv2.drawContours(resized, [c], -1, (0, 255, 0), 2)
        #print("area : "+str(area))
        #print('\nContours: ' + str(c[0]))
        #img[c[0]]
        pixelpoints = np.transpose(np.nonzero(c))
        #print('\pixelpoints: ' + str(pixelpoints))

        #  accessed the center of the contour using the followi
        M = cv2.moments(c)
        if M["m00"] != 0:
            cX = int((M["m10"] / M["m00"]) * ratio)
            cY = int((M["m01"] / M["m00"]) * ratio)
            #print (cX,cY)

            cord = img[int(cX)+3,int(cY)+3]
            print(cord)


cv2.imshow("Output", resized)
cv2.waitKey(0)
exit()

当我检查轮廓的质心颜色时，无法获取正确的颜色。有没有人知道如何使用OpenCv和Python获取轮廓内部的颜色？

- Seenu69

---

使用KMeans算法获取由轮廓指定的区域中的主要颜色。 - zindarod

我可以建议您获取每个轮廓的边界框以获得物体的四个角落。然后，我认为内部像素颜色提取很容易。 - Gowthaman

@zindarod，你能写一个答案吗？这样我就知道该怎么继续了。 - Seenu69

@Seenu69 这篇文章很好地解释了如何使用K均值聚类算法来提取图像中的主要颜色。 - zindarod

你也可以调查一下cv2.connectedComponents和cv2.connectedComponentsWithStats。也许这个链接https://www.programcreek.com/python/example/89340/cv2.connectedComponentsWithStats可以帮助你。 - John_Sharp1318

Answer 1

我简化了你的代码，并成功地获取了质心的颜色，而不使用moment。

import imutils
import cv2
import numpy as np
import matplotlib.pyplot as plt


img = cv2.imread("multiplegrains.png")
resized = imutils.resize(img, width=900)
ratio = img.shape[0] / float(resized.shape[0])
gray = cv2.cvtColor(resized, cv2.COLOR_BGR2GRAY)

ret, thresh = cv2.threshold(gray, 0, 255, cv2.THRESH_BINARY | cv2.THRESH_OTSU)
_, cnts, _ = cv2.findContours(thresh.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

# if you want cv2.contourArea >1, you can just comment line bellow
cnts = np.array(cnts)[[cv2.contourArea(c)>10 for c in cnts]]
grains = [np.int0(cv2.boxPoints(cv2.minAreaRect(c))) for c in cnts]
centroids =[(grain[2][1]-(grain[2][1]-grain[0][1])//2, grain[2][0]-(grain[2][0]-grain[0][0])//2) for grain in grains]

colors = [resized[centroid] for centroid in centroids]

- Fred Guth