通过OpenCV使用实时三相机进纸时出现GStreamer-CRITICAL错误

Question

I am using OpenCV to read live video feed from three usb cameras at the same time. This live video feed is parsed frame-by-frame into a neural network for a drone detection device (using a neural network with OpenVINO by Intel).

When using only a single camera, the following warning pops up on the screen, but the program seems to still execute flawlessly:

(python3:4235): GStreamer-CRITICAL **: gst_element_get_state: assertion 'GST_IS_ELEMENT (element)' failed

However, when utilizing all three cameras at the same time, the following errors appear in the terminal and the program does not run at all:

(python3:4041): GStreamer-CRITICAL **: gst_element_get_state: assertion 'GST_IS_ELEMENT (element)' failed VIDEOIO ERROR: V4L: can't open camera by index 0 threeCam_droneDetection.py:90: DeprecationWarning: from_ir() method of IENetwork is deprecated. Please use IENetwork class constructor to create valid IENetwork instance net = IENetwork.from_ir(model=model_xml, weights=model_bin) A camera is not working

(python3:4041): GStreamer-CRITICAL **: gst_element_get_state: assertion 'GST_IS_ELEMENT (element)' failed

(python3:4041): GStreamer-CRITICAL **: gst_element_get_state: assertion 'GST_IS_ELEMENT (element)' failed

I have tried playing around with the code in attempt to fix the problem, but this GStreamer error keeps coming up when utilizing all three cameras in the program. I even made a separate program to simply bring up a live camera feed (without doing anything with the frames except just outputting them to the screen with cv2.imshow() and this program executes flawlessly. So I assume that there is something wrong that I am not catching in the drone detection program.

The following are snippets from the python script I wrote for the three-camera live drone detection:

#Initialize camera frame dimension variables
camera_width = 244
camera_height = 244

#Initialize 3 cameras and set their frame dimensions
cap = cv2.VideoCapture(0)
cap.set(cv2.CAP_PROP_FRAME_WIDTH, camera_width)
cap.set(cv2.CAP_PROP_FRAME_HEIGHT, camera_height)

cap2 = cv2.VideoCapture(1)
cap2.set(cv2.CAP_PROP_FRAME_WIDTH, camera_width)
cap2.set(cv2.CAP_PROP_FRAME_HEIGHT, camera_height)

cap3 = cv2.VideoCapture(2)
cap3.set(cv2.CAP_PROP_FRAME_WIDTH, camera_width)
cap3.set(cv2.CAP_PROP_FRAME_HEIGHT, camera_height)

while True:
    #Read camera frames and return value (sees if cameras are working)
    ret, liveframe = cap.read()
    ret2, liveframe2 = cap2.read()
    ret3, liveframe3 = cap3.read()

    #Breaks script if any of the cameras are not working
    if (not ret or not ret2 or not ret3):
        print("A camera is not working")
        break

 # Run inference
    res = exec_net.infer(inputs={input_blob: processedImg})
    res2 = exec_net.infer(inputs={input_blob: processedImg2})
    res3 = exec_net.infer(inputs={input_blob: processedImg3})

    # Access the results and get the index of the highest confidence score
    output_node_name = list(res.keys())[0]
    res = res[output_node_name]

    output_node_name2 = list(res2.keys())[0]
    res2 = res2[output_node_name2]

    output_node_name3 = list(res3.keys())[0]
    res3 = res3[output_node_name3]

    # Predicted class index.
    idx = np.argsort(res[0])[-1]
    idx2 = np.argsort(res2[0])[-1]
    idx3 = np.argsort(res3[0])[-1]

I would really appreciate some suggestions of how I could go about removing these GStreamer errors and getting the cameras to work. Thank you!

Answer 1

Can you check the incoming video streams? You can find more info at http://answers.opencv.org/question/199105/videoio-error-v4l-cant-open-camera-by-index-0/