'A distributed solution for dynamic evidential reasoning applied to the classification of hyperspectral images'

摘要

This paper describes a distributed solution for a novel method of classification based on the evidential reasoning theory and on the implementation presented by Peddle. DER (Dynamic Evidential Reasoning) is a classification method that allows the integration of different data from multiple sources. It gives the possibility to incorporate new evidence for the classifier in order to increment its accuracy, and it also defines a different decision rule. This method is applied to the classification of different kind of crops in hyperspectral images. The images have a dimension of 1024x 600 pixels and around 10 bands per image. Different alternatives to distribute the solution were analyzed and one was implemented. The distributed solution works dividing the image to be classified among different processes which give their partial results to a master process. The solution was implemented using IDL+ENVI under a Windows PC network. Finally, experimental results of the application of the method are presented here and compared to those from the sequential solution.