The paper describes research into the use of the Belief Desire Intention (BDI) reasoning model to communicate the level of spatial uncertainty in the classification of a composite satellite image. This work involved determining suitable software agents for use in the BDI model, with a view to assisting an image analyst to better understand and manage the spatial uncertainty in their image classifications.

A case study approach was adopted that involved the use of BDI concepts to design a reasoning model that was then implemented in an agent-based prototype. Evaluation of the reasoning model and the prototype was conducted using a typical image classification scenario.

The BDI software agents provide the image analyst with a visual representation of the level of uncertainty in the composite image. This was achieved by applying BDI plan selection and plan failure and enabled the agents to generate uncertainty maps showing different views of spatial uncertainty.

This allows the analyst and the agents to collaborate on the management of the uncertainty in the image classification. The agents provide the analyst with information and feedback that can be used to refine parameters used by an empirical rule-based model to generalise noise and class uncertainty in the image classification.

This study has identified when BDI software agents can be of assistance to an analyst in understanding the level of spatial uncertainty in image analysis. Also, some insight has been gained into autonomous decision-making by the software agents in the presence of high uncertainty levels.