KBVision Constraint System Module Page 3
KBVision Module
Feature Discriminant Construction
The objective is to design a discriminant function that will filter only Bush Tokens from the Tokenset. The Token Features (Attributes) chosen for rule construction are: Excess Blue Mean, Excess Red Mean, Excess Green Mean, Compactness, Centroid Y and Pixel Count.
Each of the following six Constraint Examiner Panes show two data sets:
- A Histogram of the Feature with the objects color-coded and a function drawn by the user.
- The Tokenset filled with a gray-level equal to the value of the function as assigned by the user. Red represents a full veto of the Tokens.
The histograms of the three color features have many objects color-coded. This was done because the jacket worn by the person in the Image was much like a Bush in its visual properties. Therefore, the functions were created specifically to eliminate the region representing the jacket of the person.
The remaining histograms only have the Bush objects color-coded.

CONSTRAINT SYSTEM Page 3 of 4
Go to next Page
KBVision Modules
webmaster@aai.com (Last Update 4/19/95)
KBVision and KBView are Trademarks of Amerinex Artificial Intelligence, Inc.