Nonparametric Classification Using Parallel Systems (nsd2)
EMSL Project ID
2347
Abstract
We will use a parallel computing system to classify an image nonparametrically and compare the results to those generated by a conventional parametric supervised classification. Hopefully the nonparametric classification will be an improvement. To accomplish this task, we will be using a program written in C++ which divides each image into individual pixels, and assigns each pixel a six-dimensional position according to six attributes. It then compares that position's average distance from pixels in sample sites chosen to represent various types of artifacts to be found in the picture. The program will classify that pixel as whatever type it is closest to in 6-D. Due to the incredible number of computations required for this task, we will need a supercomputer; nothing else could perform this task in a meaningful amount of time.
Project Details
Project type
Exploratory Research
Start Date
2001-09-11
End Date
2001-10-01
Status
Closed
Released Data Link
Team
Principal Investigator