Motion detection: accumulated background or frame-by-frame?

[jrmymllr 0]

My VMS (Xeoma) has a motion detect setting where you can choose to detect differences between frames, or the accumulated background. What each method fundamentally does makes sense to me, but I was wondering if anyone has good experience/knowledge of the pros and cons of each method. Right now I'm using accumulated background which seems to work ok, but never got around to doing an in depth study.

[atcsam 0]

Accumulated background helps reduce false motion events, light rain etc..

[jrmymllr 0]

Accumulated background helps reduce false motion events, light rain etc..

 

Thanks for the reply, and sorry for my late reply. It's amazing how little information there is out there about this. What you said makes sense, but there must be reasons to use frame-by-frame. I'm using accumulated background now, also known as background subtraction, and it seems fine but have not yet fully investigated the frame differences method.

[atcsam 0]

Primarily advantage of frame by frame requires less cpu resources, especially if using keyframes only for analysis. Using a frame by frame comparison with high sensitivity can detect smaller objects in general, including shadows, cloud movement reflecting off glass, frog hopping across the driveway etc.

 

You might want to contact support to get more information

Most mfgs keep the details of their analytic algorithms close at hand and just refer to general terms of sensitivity, percentage of area, zones, and the like.

 

Look up video analytics, or video content analysis (VCA), plenty of google hits.

 

Personally I prefer edge detection and find Bosch's IVA to be one of the better implementations compared to the others.

[jrmymllr 0]

Primarily advantage of frame by frame requires less cpu resources, especially if using keyframes only for analysis. Using a frame by frame comparison with high sensitivity can detect smaller objects in general, including shadows, cloud movement reflecting off glass, frog hopping across the driveway etc.

 

You might want to contact support to get more information

Most mfgs keep the details of their analytic algorithms close at hand and just refer to general terms of sensitivity, percentage of area, zones, and the like.

 

Look up video analytics, or video content analysis (VCA), plenty of google hits.

 

Personally I prefer edge detection and find Bosch's IVA to be one of the better implementations compared to the others.

 

I was using accumulated background for a few weeks and was getting frustrated with it. I assumed it was better, but it seemed to pick up too many false events, didn't catch the ones it should have, and had trouble at night. Additionally, having the motion detection ignore any movement less than 0.5s, something I thought would be useful, didn't seem to work well with accumulated; turning on that feature made the motion detector miss nearly everything unless the sensitivity was turned way up.

 

Since going back to frame-by-frame and ignoring events less than 0.5s, things seem to be working better so far. The sensitivity doesn't have to be turned up quite so much, it's not triggering on every rain drop, and motion at a distance is detected fairly easily.

 

Will try google searching with the keywords you mentioned.