pixel change detection [closed]

asked 2018-06-13 20:12:36 -0500


Fair warning, total noob question. I'm curious to see if OpenCV is even the correct library for this use case.

The user would like to zoom into a section of the footage, an "area of interest" to quickly see all the "clips" of when any activity took place within said "area of interest".

On a primitive level, this is essentially pixel change in a localized area, so I'm not sure machine learning or CV is necessary - but I'm open to whatever the simplest solution is, especially if something out of the box already works.

But ideally, we'd like to train a model so it knows what is "normal" activity (i.e. mailman), and what isn't (i.e. vandal).

Any feedback on the simplest way to make this happen is greatly appreciated!

Thanks, Nick

edit retag flag offensive reopen merge delete

Closed for the following reason question is off-topic or not relevant by berak
close date 2018-06-14 01:43:50.245335


what you want is called "action recognition" or "abnormality detection", there is nothing like that readymade, and you would need real skill to implement it.

asking around won't help you, please have a look at the faq

again, this is an active research topic, not a "solved problem". given your lack of reseach and skill, you won't be able to do anything here.

berak gravatar imageberak ( 2018-06-14 01:42:18 -0500 )edit