Naked Science Forum
Non Life Sciences => Technology => Topic started by: thedoc on 15/06/2012 02:30:02
-
Jon Kelling asked the Naked Scientists:
Dear TheNakedScientists,
Is it possible to design a basic nondigital motion detector that is sensitive to shape and/or color in addition to light?
If so, what principles would be involved in making this happen?
Thank you,
Jon Kelling
Madison, WI
USA
What do you think?
-
I guess it would involve grids of lasers and measuring distances of reflection. Let's say you took the shape you are trying to detect, put points on it in 3D space. Now put a virtual motion detector some X distance from this shape and place it directly in front of the shape. Now map whatever points on the shape that are visible to the detector from this position on a grid as well as the distance the points are from the detector.
Repeat this procedure by moving the detector in a circle around the shape at set intervals, while recording the points that are visible for each position on a grid. Overlay all these points for each position on a single grid, while keeping each point mapped to its relative position to the shape.
Use this grid to create an actual grid of lasers, spin it around like you would in orb barcode scanner so you can normalize orientation, and you should be able to recognize when a particular object is matched.
Being non-digital, of course, such a technique would leave the object you can match essentially "hard-wired" to the circuitry.
-
https://en.wikipedia.org/wiki/Lidar ?
[ why the aversion to digital electronics ?, an analog electronic system would use transistors, which are used to make digital electronics ]
-
And perhaps a good way to "double-check" your results would be boundary detection. By that I mean at each position, create arbitrarily positioned points just outside the 2D boundary created by the shape. When these points are mapped to lasers, then you can be sure that, for example, it's not a relief of what looks like the shape you're trying to match on some bigger object, but rather the actual object since it's boundary is also matched correctly.
So you will have matched the 3d shape of the object, plus the boundary of the object at a particular position. And you could just as well rotate the detector in a 3D sphere around the object when recording points if you needed it to work in truly any position relative to the object you're trying to match. You would of course need to pack lots and lots of lasers in close proximity the more flexible you want to make your system.
The only reason I can imagine for using an analog system is cost. I was thinking along the lines of barcode scanners while writing this fun thought experiment (standard laser barcode scanners are typically cheaper, sometimes far cheaper than digital "imagers" that can scan both 1D and 2D codes).
-
i would think a combination of motion detector technology with facial detection software, in which you would enter the shape you want to detect instead of the face, would work. theoretically anyway.