Ask Your Question

Delay when grabbing frames from webcam

asked 2016-03-26 09:46:27 -0500

Klee gravatar image

I want to use an external trigger to take a single photo from a webcam. However, my webcam behave differently than I expected. I placed the webcam in front of a digital clock and took photos every 2 minutes and then saved the photo as .jpg with the filename containing the current timestamp. The time on the first photo is the same as the first timestamp. Then the next four photos also show the clock showing the same time as the first timestamp (although they were taken at 2 min, 4 min, 6 min and 8 min). The clock on the 6th photo shows the time of the second timestamp (clock shows 2 min whereas jpg was saved at 10 min), the 7th photo shows the time of the 3rd timestamp and so on.

Timestamp jpg     Clock in the photo
00:00                   00:00
00:02                   00:00
00:04                   00:00
00:06                   00:00
00:08                   00:00
00:10                   00:02
00:12                   00:04
00:14                   00:06
00:16                   00:08

I have tried 2 webcams, 2 operating systems, and different time intervalls between 10 seconds and 2 minutes. I have tried the opencv function "read" as well as "grab" and "retrieve" and see no difference.


import cv2
import time
cap = cv2.VideoCapture(0)
steps = [120,120,120,10,10,10,60,60,10,10,10,10,10,10,0]
for n in range(len(steps)):
    t = time.strftime("%Y-%m-%d_%H-%M-%S")
    #ret, frame =
    ret = cap.grab()
    ret,frame = cap.retrieve()
    cv2.imwrite('test_' + t + '.jpg',frame)

Is this the normal expected behaviour of a webcam or is something broken here? Am I using the right functions for my purpose?

Thank you for your help!

edit retag flag offensive close merge delete


My first guess, like @berak said, you are using webcams with an internal frame buffer. Most of those provide you with an own SDK that allows you to set the buffer to contain only a single frame. By doing that you will always get the latest frame. A quick and dirty solution is to always grab 5 frames at a time and only store the latest one (an average frame buffer is about 2-5 frames in size).

StevenPuttemans gravatar imageStevenPuttemans ( 2016-03-29 04:40:46 -0500 )edit

3 answers

Sort by ยป oldest newest most voted

answered 2016-03-26 21:21:47 -0500

berak gravatar image

updated 2016-03-26 21:29:15 -0500

some VideoCapture implementations (i.e. v4l) have an internal ringbuffer for the images. it seems, you're not flushing that fast enough, and so you get outdated images.

try to change your logic, read images continuously from the cam (don't sleep that long), and save, when appropriate

import cv2
import time
cap = cv2.VideoCapture(0)
steps = [120,120,120,10,10,10,60,60,10,10,10,10,10,10,0]
n = 0
next = time.time() + steps[0]
while True:
    ret, frame =
    if not ret : break
    now = time.time()
    if now >= next:
         n += 1
         if n >= len(steps) : break
         next = now + steps[n]
         t = time.strftime("%Y-%m-%d_%H-%M-%S")
         cv2.imwrite('test_' + t + '.jpg',frame)

(note: unfortunately no python on this box, so this is untested)

edit flag offensive delete link more

answered 2020-01-02 10:17:25 -0500

arnthorg gravatar image

I know this is old but maybe setting the buffer size would help

self.cap.set(cv2.CAP_PROP_BUFFERSIZE, 1)
edit flag offensive delete link more

answered 2016-03-30 02:29:49 -0500

Klee gravatar image

Thank you. I think I will not be able to read images continuously, because in my real application, my python code is doing lots of other stuff and there is no sleep time between two images. But I will always read in 5 or 10 images and throw away all but the last.

edit flag offensive delete link more

Question Tools

1 follower


Asked: 2016-03-26 09:46:27 -0500

Seen: 6,913 times

Last updated: Jan 02 '20