Python 2.7 / Opencv3.0：使用cv2.connectedComponentsWithStats时质心数不正确。

Question

我在Python 2.7中使用cv2.connectedComponentsWithStats来查找二进制图像中三个blob的质心。 （图片在In code it is img3下面给出）。

在此图像中，存在三个白色bolb，因此质心的数量应仅为三个，但python显示8个质心值

[[ 307.08579803   95.31441513]
 [ 511.51325364  141.06288981]
 [ 259.          112.5       ]
 [ 296.07169811  129.18490566]
 [ 276.          140.5       ]
 [ 225.19643047  237.16190375]
 [ 189.21212121  271.6969697 ]
 [ 187.58333333  285.83333333]]

每个bolbs的面积在最后一栏中给出

[[     0      0    640    480 286024]
 [   260     65    100     68   3753]
 [   454     84    105    118   7696]
 [   259    112      1      2      2]
 [   277    117     34     23    265]
 [   276    140      1      2      2]
 [   168    173    114    128   9413]
 [   186    268      8      8     33]
 [   186    284      5      4     12]]

我的代码如下：

import numpy as np
import cv2
from PIL import Image
import matplotlib.pyplot as plt
import timeit
from skimage import morphology, measure

img = cv2.imread('left180.png')

lower_white = np.array([150,150,150], dtype=np.uint8)
upper_white = np.array([255,255,255], dtype=np.uint8)

binimg = cv2.inRange(img, lower_white, upper_white)
cv2.imshow('res',binimg)

binimg[binimg!=0] = 255
# flood fill background to find inner holes
holes = binimg.copy()
retval, image, mask, rect = cv2.floodFill(holes, None, (0, 0), 255)
# invert holes mask, bitwise or with img fill in holes
holes = cv2.bitwise_not(holes)
#cv2.imshow('holes',holes)
filled_holes = cv2.bitwise_or(binimg, holes)
cv2.imshow('filled holes', filled_holes)

imglab = morphology.label(filled_holes)
cv2.imshow('label',imglab)
cleaned = morphology.remove_small_objects(imglab, min_size=1264, connectivity=4)
#np.savetxt('cleaned.csv',cleaned,fmt='%.18g',delimiter=',')
cv2.imshow('clea',cleaned)
img3 = np.zeros((imglab.shape))

img3[cleaned > 0] = 255
img3= np.uint8(img3)  #### here conversion of array into uint8 data conversion is important
                      ### else cv2.connectedComponentsWithStats will show error.
#np.savetxt('img3.csv',img3,fmt='%.18g',delimiter=',')
nb_components, output, stats, centroids = cv2.connectedComponentsWithStats(img3, connectivity=4)
sizes = stats[1:, -1];
#nb_components = nb_components - 1
cv2.imshow("centroid", img3)
centroids1 = centroids[1:nb_components]
print(centroids1)
print(stats)
cv2.waitKey(0)

为什么会出现这个问题？我想获得blob的质心，其面积大于1000像素。怎么做？