Date: Wed, 24 May 2006 23:43:35 -0500 (CDT)
Subject: opencv cross compiled
X-UID: 174

OpenCV (open computer vision) is cross compiled into the uClibC/buildroot 
environment. It was faster to copy and sed the configure generated files 
rather than properly modify the build system. I will have to clean up the 
toolchain and system software that goes on the robot. However, that can 
wait until later.

I'm trying to follow the principle of working as if someone were paying 
me. In this case, the prime directive would be, "Make it work quickly!" 
Neat and tidy would be secondary. So far, I've never seen any place that 
valued cleanliness over expediency and function, despite what might be 
claimed.

I found an excellent description of the Mahalanobis norm.

http://diuf.unifr.ch/courses/05-06/patrec/annexes/Mahalanobis.pdf

This is the statistical distance. From the first page, there is a 
geometric description. Conceptually, Euclidean distance is rescaled in 
directions for the corresponding random variables according to the level 
of statistical uncertainty.

When I was an undergraduate, I would have tried to bull my way through 
lots of algebra using the left side of my brain. Nowadays, I can not 
understand anything without geometric visualization like this. An 
advantage is greater intuition about how things work.