Cam prefilters 200pics/d - opencv able to detect 'foreign objects' within those?

asked 2016-01-24 18:19:03 -0600

itsmee gravatar image

updated 2018-01-10 07:15:51 -0600

Hi guys,

more of a fun project than something serious - but pretty curious.

I compiled opencv 3.1.0 on Banana Pi yesterday night & was able to bring up reasonable results with haarcascade_frontalface as well as "HOG descriptor/person detector". Impressive framework.

Exercise:
- Cam in my frontyard has motion detection ~150-300 pics per day shoot last week
- Pics are from a static area - if nobody is in very boring & VERY similar
- Would like to filter down to 10-20 with objects inside. (opencv filter down to 50 which are not blank & than apply custom logic 'only one pic per 30 secs'. Will still have all if details on an object are required).

Now my thinking is - its cool to detect faces or persons. But it also sounds rather complex. Is there an opencv way to correlate the pictures to identify objects present in one and not in the other? Wouldn't care if it gives a false alarm for the cat.

Thanks for reading, guess the answer is 'not really' - but maybe you know more than google told me :-)

ItsMee

edit retag flag offensive close merge delete

Comments

1

not sure, if i understand you correctly, - would background subtraction help ? like, only take a shot, if something is moving.

berak gravatar imageberak ( 2016-01-24 20:42:39 -0600 )edit