Untitled

Information Engineering - I0n

ASSIGNMENT DETAIL Back

Insert Date:	3/15/2005
Date Assigned:	3/15/2005
Date Due:	3/16/2005
Assignment:	2D FFT tutorials

Please see the links below. You should be able to use these tutorials to get an idea of how the 2D FFT can be used to quickly identify objects.

I don't expect you to be able to do the double integrals, but instead use an engineering approach to these VAB blocks. Can you find a way to use a portion of a scanned objects' FFT and then compare it to a known object FFT (we'll control size and brightness parameters). The question to be answered then is:

1. How many frequencies in a 2D FFT are required to accurately identify an object that the camera is seeing? How robust is this method?

2. Can multiple FFT signatures from multiple objects within a single scan frame allow multiple recognitions?

> The Beobot team uses HST (hue, saturation, value) in order to identify salient features.

Assignment Links:

	2nd tutorial: Graphical interpretation of the FFT Taking colored frequency plots to track where the information is stored in the resulting matrices (even and odd harmonics). Written by U of Toronto EE Prof (who also is part cyborg it appears...).
	3rd tutorial: Neuromorphic Engineering Newsletter Volume 1, Issue 1 from Spring of 2004. Read pages 10, 12. The rest is just plain cutting edge! Warning: a 2.4 MB pdf file!
	Math descriptions of the FFT Useful for the Calculus folks...
	1st tutorial: Picture Book tutorial of 2D FFT's Kevin Cowtan's tutorial was written with crystallography in mind, but is a great visual introduction (or should I say introDUCKtion) to the topic. Hew orks at U. of York's Structural Biology Laboratory.