Key words: Expert Systems, Remote Sensing, Conflict Resolution, Image Classification
This paper investigates the potential benefits an inference capability can provide for automated image classification tasks in remote sensing. Most current applications depend upon skilled image interpreters to provide adequate classifications of satellite images. This creates problems in that it takes time to classify images manually, skilled interpreters may not always be available, and often, manual classification can be subjective (dependent on the interpreter). This paper addresses automating the classification process using knowledge-based or expert system technology in an attempt to solve these problems.
KAGES (Knowledge Acquisition for Geographical Expert Systems) is a knowledge acquisition system designed to obtain knowledge from image interpretation experts. It allows the user to visually and non-visually classify objects in remotely sensed images. KAGES produces rules for identifying points, lines, and polygons (two-dimensional features). This paper's focus is on polygon rules and excludes rules identifying points and lines. Rules already produced through the knowledge acquisition process in KAGES are used. An inference capability for this knowledge acquisition system involved taking the CLIPS rules and applying them to remotely sensed images. The remotely sensed images were obtained from the SPOT HRV (Systeme Probatoire d'Observation de la Terre, Haute Resolution Visible) satellite.
One of the main benefits that an inference capability can provide to remote sensing is its ability to resolve conflicts. Conflict can occur in classifying satellite images when a pixel in one band is assigned a label such as trees, and the same pixel on another band is assigned to another type of object, say water. Without the help of a skilled interpreter it is difficult to classify this pixel. Most spatial information systems and image processing packages do not provide the reasoning capabilities needed to make such decisions. Expert system technology provide many different conflict resolution strategies to help make these decisions.
This paper includes the implementation of three conflict resolution strategies. These strategies include first in first serve, last in first serve, and prioritisation. Prioritisation in particular is an important technique for resolving conflicts, as the setting of priority on rules requires the knowledge of a domain expert. Other conflict resolution strategies such as specificity and recency are also discussed.
One other important aspect addressed in this paper is the time the system takes to produce a classified result. CLIPS is unable to process whole images and needs to do the processing pixel by pixel. Pixel by pixel reasoning can slow down processing noticeably. Each SPOT image is 894 x 896 pixels multiplied by three, because SPOT images include three bands. This paper includes investigation into different methods of implementation to solve the problem of slow processing.
The integration of remote sensing and knowledge-based systems that this paper describes provides solutions to the problems of the large amounts of time needed for manual classification, the availability of skilled interpreters, and the inherent subjectivity of manual classifications. The integration of an inference engine and conflict resolution will help speed classification and does not require interpreters to always be available.