FuRII: An ENVI classification toolbox – Evaluation and validation


  1. Leduc, F.
Corporate Authors
Defence R&D Canada - Valcartier, Valcartier QUE (CAN)
This document is the last of a trilogy related to FuRII, a per-pixel image classification tool developed at DRDC Valcartier and integrated as a toolbox into ENVI, an image-processing tool developed by Exelis VIS. FuRII (Fuzzy Reasoning applied to Image Intelligence) is an information fusion tool in which the imprecise knowledge on objects to extract from images is modeled with membership functions. Fusion may be performed within the framework of the fuzzy set theory or the evidence theory. In the latter case, it is possible to select a closed-word or an open-world paradigm. In all cases, it is possible to weight sources according to their reliability. Within FuRII, objects to extract are modeled independently in each image component (spectral bands, texture images, etc.) thus allowing the ingestion of multiple source data. Results showed that the integration of reliability in the fusion process did not allow significant improvement in the classification performance. It is also demonstrated that the simplest fusion operators perform better. Among them, the fuzzy quantified adaptive fusion is identified as the best operator to use in terms of performance and computation time. FuRII’s results have also been compared to that of the traditional maximum likelihood classifier (MLC). It is found that the MLC performs better when data is normally distributed. On the opposite, with non-normal distributions, quantified adaptive fusion operator performs better. Although FuRII is design

Il y a un résumé en français ici.

FuRII;image classification;multisource fusion;fuzzy sets;evidence theory;Dempster-Shafer;membership functions
Report Number
DRDC-VALCARTIER-TR-2012-132 — Technical Report
Date of publication
01 Apr 2013
Number of Pages
Electronic Document(PDF)

Permanent link

Document 1 of 1

Date modified: