Combining Supervised Remote Sensing Image Classifiers Based on Individual Class Performances
This article focuses on the use of multiple classifier systems (MCSs) based on dynamic classifier selection. Four implementation strategies of MCSs are compared: majority voting, belief networks, and two designs based on dynamic classifier selection. Experimental results indicate that the direction taken by Woods et al. is the best alternative for remote sensing applications for which the classifier- dependent posterior distributions are unknown.
Bibliographic Reference: An oral report given at: 2nd International Workshop on Multiple Classifier Systems (MCS 01). Organised by: University of Surrey and Robinson College. Held in: Cambridge (UK), 2-4 July 2001
Record Number: 200013532 / Last updated on: 2001-07-25
Original language: en
Available languages: en