Backprojection not working when tried with BGR instead of HSV

edit

Are you sure that 's OK in HSV? @berak i think there is a bug in opencv-python when you give three channels

My program is

import numpy as np
import cv2 as cv
img = cv.imread('c:/lib/opencv/samples/data/fruits.jpg',cv.IMREAD_COLOR)
orange = img[300:380,150:220]
cv.imshow('fruits',img)
cv.imshow('orange',orange)
# calculating object histogram
c=[0,1,2]
r=[32,32,32]
rc=[  0, 256,0,256,0,256]
roihist = cv.calcHist([orange],c, None, r, rc )
# normalize histogram and apply backprojection
cv.normalize(roihist,roihist,0,255,cv.NORM_MINMAX)
dst = cv.calcBackProject([img],c,roihist,rc,1)
ret,thresh = cv.threshold(dst,10,255,0)
thresh = cv.merge((thresh,thresh,thresh))
cv.imshow('dst',dst)
cv.imshow('res',thresh)
cv.waitKey(0)
cv.destroyAllWindows()

thresh image is black. In c++ I have got a good result.

when I set

 c=[0,2]
 r=[32,32]
 rc=[  0,256,0,256]

result is good. I don't understand what is wrong in 3D

ok, i can reproduce it. (but no idea, so far)

I think problem is https://github.com/opencv/opencv/blob... . I don't call this function in C++ I call https://docs.opencv.org/master/d6/dc7...