MD camera has only one .unf motion capture, but triggered camera has dozens?
  • I'm looking for a bit of clarity on how the triggering/retriggering occurs.

    I have a camera with two streams: a 320x240 low res stream, and a 1920x1080 high res stream. I have these streams entered as two cameras in SS, namely "Garage Front MD" and "Garage Front".

    The MD camera is set up for motion detection. 100% sensitivity, 0.5s trigger time, 75% human/vehicle for both motion detection and for trigger actions. No audio. This MD camera is (temporarily) set up to capture a movie in response to motion trigger. One movie per event, 10fps, 15s pre capture, 30s post capture. Its actions are set up to trigger the "Garage Front" camera, highlight the camera with red box for 5 seconds, and 5s action reset time.

    The triggered camera ("Garage Front"), does NOT have motion detection turned on in the "Triggers" tab, but does have "Capture movie in response to motion trigger" enabled, one movie per event, 30fps, 15s pre capture, 30s post capture. It has no actions enabled, and has the same 5s action reset time.

    So right now (and for the past hour or so since enabling MD's capture video on motion detection, I have one large, growing .unf file for the MD camera, but dozens of short (45-50s) high-res videos from the triggered camera. I am assuming the latter is because the MD camera is detecting motion and sending trigger events to the high res camera, and those trigger events are far enough apart (15s pre, 30s post, so 45s between events) to count as separate events. This doesn't explain why the MD camera motion capture video is one long continuous file that in over an hour has not stopped (it's still named ".unf").

    Watching the status bits in the "Camera Info" window shows the MD camera motion 100% almost all the time with "M" red. The triggered camera shows M/A red periodically, with M on considerably longer than A, and of course no motion detection since it's not enabled. I suspect the 100% motion detect almost all the time on the MD camera has something to do with all the "MOTION 0 0 0 0" events I see. Perhaps in a future update the Camera Info window might show the AI results for Human and Vehicle too as a bar graph like the Motion field, that'd help for tuning as well.

    Back to my thread question though: how is the self-trigger of a motion capture event different from an other-camera-trigger motion capture event?
  • The most obvious thing is the "Capture type" setting under Preferences -> Cameras -> Motion detection. Are you sure they are both set to "One movie per event"?

    If this is not the issue, then your results do seem inconsistent with your settings. I'll do some testing here to see if we can reproduce this issue and find any problems with the mechanism of one camera triggering another.

    Are any recordings on the high-res camera significantly more than 45s in length?

    My other comment is that with such a high sensitivity and so much motion, near-continuous capture is likely, and you may as well use continuous-capture recording rather than motion-detection recording.
  • yep they are definitely identical in that setting; in fact the only difference between them on that tab is 30fps for HD vs 10fps for MD.

    There are *some* HD videos which are longer than 45s, but not many during this kind of activity.

    I agree that 320x240 (not even 640x480) is probably not going to work real well (especially at night) for AI. It's too bad we can't look at the low-bitrate, low-res stream for motion detection and then give it a snapshot from the high res stream to do the AI on. Unfortunately continuously decoding frames from a 1920x1080 stream for MD is too resource-intensive, and even if we could grab a frame from the HD stream for AI, there's a not-insignificant amount of work to "roll back" the h.264 HD stream to find an I-frame to give to the AI.

Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!