An APL rule-based system architecture for image interpretation strategies

Abstract

APL is both a language of choice for image processing and image description tasks, and a language for artificial intelligence applications, typically expert systems. The paper presents the architecture of an APL rule-based system for realizing description synthesis strategies i.e., sequences of actions which allow the evaluation and interpretation of the properties of structures detected in an image. Descriptions of images are stored in a data structure which is formally presented in APL2 syntax. Describing an image often requires that reflective actions be taken i.e., actions in which the system examines the state of the computation and its internal state to select the next action to be executed. It is shown how APL provides all the features needed to implement a reflective mechanism through the use of metarules of the same format as rules.