Video processing... software etc..

StickyMcGinty

Lads,

Is it possible to monitor a stream of video and set a trigger when something appears on the stream i.e. have a video stream from a room so when something changes it triggers an event.

any ideas on what kind of software/hardware would be needed for this?

Cheers

aidan_walsh

Depends on the type of change you are looking for.

By and large, this is how CCTV systems record information, by monitoring the data coming in fromt eh camera, checking histogram changes and switching to a different monitoring mode of there is a drastic change, such as a person entering the room, or a dark room suddenly becoming bright.

StickyMcGinty

cheers aidan

I'm looking to flag more of a position/perspective change, like when the camera sees a certain side of a rotating object then it triggers an event. (this side of the object will have something distinguishing that the camera can recognise, such as a reflective strip)

ressem

Depends on a lot of stuff.
Whether you need hardware depends on how fast you need to process the images and what cameras/video input you will be using. Is it for industrial use or a student lab project.

As for software, will the camera be moving? What's the background, is it reflective, does it change and the likelyhood of false positives?

If the side is reflective/highly coloured side and illuminated, and the background is fixed, then a histogram might do. Though you might have to might have to compensate for colours introduced by video compression and other artifacts.

If you need more complicated object recognition you might want to look at the tutorials at http://www.roborealm.com/tutorials.php

and libraries
http://www.mathtools.net/C_C__/Image_Processing/index.html

StickyMcGinty

cheers ressem great to hear some feedback

yea its for a student lab project so there is a small enough budget. The camera will be stationary it'll only be a single object that will be moving. the background will be dark and wont change, do you reckon attatching something like a reflective strip to the object would be the best way to achieve a consistently accurate positive result?

stevenmu

I've never done anything like this, but I'd imagine a different colour would be slightly better than a reflective strip. Either something like a bright white, so you could look for overall brightness changes, or possibly a particular shade of blue or green that you can scan the images for.