Friday, March 16, 2012

Light Fields

We seem to be living at a time of innovation in sensor technology.  Light field cameras are another possible alternative to either a Kinect type sensor or conventional stereo vision.

Since service robotics is likely to be mainly concerned with objects within a relatively short distance of a few metres, something like this could be viable depth sensor.  Compared to a Kinect it would have a lower cost both in terms of price, due to the lower number of components, and also in terms of energy (and hence battery life), since it doesn't depend upon pumping out a lot of energy into the environment.

1 comment:

beambot said...

IMO, Lytro is a good example of the vision techniques generally referred to as Computational Cameras. Actually, these methods have been around for quite some time... as had the basic idea behind the Kinect.

So... what's special about right now that makes these ideas commercially viable? The best I can come up with... the video game industry trying out new paradigms (Kinect) and fortuitous VC funding (for Lytro). Is there something more to this? Is there something we could learn from these examples to make it happen again (and again, and again)?

Either way, the low-cost sensors are a big deal for robotics. Next up... actuators!!