electron's profile - activity

Maybe you are right but the main problem here to convert to right opencv format

I have PyQt5 QPixmap. I want to convert it to the same format as cv2.VideoCapture returns. How to achieve this? I use OpenCV in Python3.

because cm count per pixel is different in these cases. distance in cm is the same, but in pixels it will change depends on your camera position. that's normal

using your code. you wrote that you can find centers of boxes and seems that your code really can do this, just save coordinates cX, cY to appropiated variables

For example, cX1 and cY1 are coordinates of center of first box, cX2 and cY2 are coordinates of center of second box. You can find the distance:

 distance = (((cX1 - cX2) ** 2) + ((cY1 - cY2) ** 2)) ** 0.5

If you know real sizes of boxes you can also compute distance in cm. But accuracy will be low.

you mean distance in pixels?

This happens when for some reason cap.read() returns empty frame. just add a check:

while True:
    ret, frame = cap.read()
    if frame:
        gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
        cv2.imshow('frame', gray)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

btw, seems that you are trying to use in Python37 opencv installed for Python34. Check if you're using correct python versin and make sure that python is in PATH

download wheel and install it using pip: pip install path_to_wheel

Are you limited in using 3.2.0? If no I suggest to try prebuilt wheels: http://www.lfd.uci.edu/~gohlke/python...

11 channels per single image? wow. seems that trivial cv2.merge() cannot do it. so probably slices?

1) Check if file exists. Try to use absolute path there, and use '/' instead of '\' if you are on windows

2) According to docs:

Sometimes, it is a headache to work with Video Capture mostly due to wrong installation of ffmpeg/gstreamer.

It can be an issue too

depends on what object you want to detect

Not sure that cv2.HoughCircles is the best option here. I recommend to try findContours also. Anyway question is about houghCircles, so example of code that detects all planets correctly:

import cv2
import numpy as np

planets = cv2.imread(path)
gray_img = cv2.cvtColor(planets, cv2.COLOR_BGR2GRAY)
gray_img = cv2.medianBlur(gray_img, 3)
circles = cv2.HoughCircles(gray_img, cv2.HOUGH_GRADIENT, 1, 10,
                    param1=350, param2=10, minRadius=2, maxRadius=13)
if circles is not None:  # circles found
    circles = np.uint16(np.around(circles))

    for i in circles[0,:]:
        cv2.circle(planets,(i[0],i[1]),i[2],(0,255,0),2)
        cv2.circle(planets,(i[0],i[1]),2,(0,0,255),3)

cv2.imshow("HoughCirlces", planets)
cv2.waitKey()
cv2.destroyAllWindows()

What I corrected in your code: set minimum distance between circles centers from 120 to 10, changed min and max radius for circle, changed param1 and param2. Also I adde check for case if no circles were found.

Docs that I used: link

Hope this helps!

could you please provide code? are you sure that file that you try to open exists? try to show it using opencv and see what happens

I know it's too late, but this range works well for me:

lower_blue = np.array([100,150,0])
upper_blue = np.array([140,255,255])

thanks! i will try it

btw i don't sure that convertion works properly because saved on disc out_image looks strange

I need YUV420, not YUV. so BGR2YUV_I420, not BGR2YUV

cv2.cv is in opencv2, not in opencv3+. just use cv2.what_do_you_want() instead of cv2.cv.what_do_you_want()

I use OpenCV 3.0 in Python3.4. I have BGR image converted to YUV420:

out_image = cv2.cvtColor(in_image, cv2.COLOR_BGR2YUV_I420)

How to get Y, U, V values from out_image?