abysmal performance of code

The code is supposed to take a series of snapshopts fom 2 individual ip cameras and save these images in 2 seperate files.

However instead of a set of pictures being taken every second or so the time in between pictures is more in the range of 5 seconds which is a bit too long for my project. If you have any idea on how to speed it up let me know.

mkdir irbilder1
mkdir norbilder1

camA = videoinput("gentl", 1, "BayerBG8"); % Modify properties based on your first camera
camB = videoinput("gentl", 2, "Mono8"); % Modify properties based on your second camera

camB.ReturnedColorspace = "rgb";

src = getselectedsource(camB);
src.AtmosphericTemperature = 223;
src.CMOSBitDepth = "bit8bit";
src.ReflectedTemperature = 223;
src.WindowTemperature = 223;
Frames Aufnehmen und ggf in Grauwert konvertieren


nummer=1;
ende=10
while nummer<=ende
snapA=getsnapshot(camA);
    snapA=im2gray(snapA);

    snapB=getsnapshot(camB);
    snapB=im2gray(snapB);
  itr=string(nummer)
    nameA="nom"+itr+".png"
    BezA=fullfile("norbilder1",nameA)

    nameB="IRB"+itr+".png"
    BezB=fullfile("irbilder1",nameB)

    imwrite(snapA,BezA);
    imwrite(snapB,BezB);
    beep
    nummer=nummer+1;
    
end

Answer 1

[Pratik]

Hi Frederik,

As per my understanding, you are trying to capture images from two cameras simultaneously and save them to a folder, but the process is too slow.

After analysing the performance of the code, the bottleneck seems to be the “getsnapshot” function. Each call to the function is taking approximately 2 seconds.

To optimise this process, it would be more efficient to record a video rather than taking individual photos. This way, one can easily extract frames at desired intervals, ensuring that the necessary images are captured.

For comparison, time taken to click 10 images by a single camera is around 24.76 seconds. By capturing frames from a video, 10 frames can be captured in ~9 seconds.

Please refer to the below code snippet to record a video using a camera and access its frames:

% Set up the video input object
v = videoinput("winvideo", 1, "YUY2_1280x720"); % modify property based on your camera
% Set the properties for the video object
% Set frames per trigger to 5 seconds worth of frames
% Assuming the default frame rate is 20 frames per second
fps = 20; % Default frames per second, change if different
duration = 5; % Duration of recording in seconds
framesPerTrigger = duration * fps;
set(v, 'FramesPerTrigger', framesPerTrigger);
% Set the trigger repeat to 0 to only run acquisition once
set(v, 'TriggerRepeat', 0);
% Set the returned color space to RGB
set(v, 'ReturnedColorSpace', 'rgb');
tic; % Start timer for this iteration

% Start the video acquisition
start(v);
% Wait until the acquisition is finished
wait(v, duration + 5);
% Retrieve the frames and timestamps
[frames, timestamps] = getdata(v, v.FramesAvailable);
% Stop the video object
stop(v);
elapsedTime = toc; % End timer for this iteration
fprintf('Time taken for iteration: %.6f seconds\n', elapsedTime);

% Extract and display a frame 
for i = 1:2*duration
    % Calculate the frame index 
    frameIndex = i * fps/2;
    % Extract the frame 
    frame = frames(:,:,:,frameIndex);  
    % the frame can be saved after preprocessing 
    % Display the frame
    imshow(frame);
    title(sprintf('Captured frame %d', i));
    % Pause for a moment to display the frame 
    pause(1); 
end
% Clean up by deleting the video object 
delete(v);

Please refer to the following documentation of ‘videoinput’ for more information:

https://www.mathworks.com/help/imaq/videoinput.html

Hope this helps!

[Frederik Vogel]

Hey Patrik,

thank you for your attempt, however i have tried this approach as well. however this system will take a video from one camera terminate the video and then record a second subsequent sequence from the second camera.