cv2.videocapture.read() does not return a numpy array

Question

I have this code trying to capture a frame from my webcam on raspberry pi, and saving it as an image. I use opencv 2, but I get strange errors when I run the code..

import time
import sys
from subprocess import call
import cv2

cam = cv2.VideoCapture()

while True:
        cam.open(-1)
        image = cam.read()
        cv2.imwrite("current.jpeg",image)
        time.sleep(10);

This is what the program returns:

VIDIOC_QUERYMENU: Invalid argument
VIDIOC_QUERYMENU: Invalid argument
VIDIOC_QUERYMENU: Invalid argument
VIDIOC_QUERYMENU: Invalid argument
VIDIOC_QUERYMENU: Invalid argument
VIDIOC_QUERYMENU: Invalid argument
VIDIOC_QUERYMENU: Invalid argument

Traceback (most recent call last):
  File "kvamskogen.py", line 18, in <module>
    cv2.imwrite("current.jpeg",image)
TypeError: <unknown> is not a numpy array

What is wrong here?

Yumuk

"You can ignore the ftp part, and the commented lines." - you should edit the irrelevant parts out then to help people who are trying to help you. Rather than a) Leaving extra code in your example and b) explaining why it isn't relevant. See sscce.org – Charest

Answer 1

Reading (cam.read()) from a VideoCapture returns a tuple (return value, image). With the first item you check wether the reading was successful, and if it was then you proceed to use the returned image.

This is documented at https://opencv-tutorial.readthedocs.io/en/latest/intro/intro.html#capture-live-video

Gnawing

@mmpg do you know what might be the reasons of a cam.read() exception? I have an app that after some period of time running starts to raise exceptions from cam.read(). It shows previews frames with .set(3, 640) and .set(4, 360) but raise the exceptions NULL object passed to Py_BuildValue with .set(3, 1920) and .set(4, 1080). cam.read() can't even return False in the tuple, it just crashes. – Electromagnetic

Answer 2

Everything mmgp said is spot-on; cam.read() returns first a boolean indicating whether the read was successful, and then the image itself (which will be empty if the return value was False). Also note that if you're not using the return value for anything, you can just set that portion to _, which tells Python "ignore me"; that line would then look something like _, image = cam.read(). Additionally, it is generally good practice to specify the index at which your camera is located (usually 0 if you have only one camera connected) when calling cv2.VideoCapture(), so that, in the event that you do have multiple cameras connected, OpenCV knows which camera to read from (otherwise it might just crash because it doesn't know what to do).

Isomerize

FWIW, calling a variable _ doesn't tell Python anything. Python just treats it as another variable name. It's merely a convention programmers use to tell themselves and others that you're not doing anything with this value. – Bennet

Answer 3

You should use arguments in cv2.VideoCapture()

Try this to capture from the default camera

cam = cv2.VideoCapture(0)

Try this to capture from ip camera

cam = cv2.VideoCapture('http://ip-address') # to check video source's ip address right click video and select "copy image address" and put the exact address in above line of code

Tobi

Answer 4

If you are trying to capture video using inbuilt webcam then following code line will give you best results cap = cv2.VideoCapture(0, cv2.CAP_DSHOW)

Standardbearer