Rule-based interpretation of aerial imagery.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Geographic Information Systems

Background:

• Interpreting complex scenes like airports requires integrating diverse data sources.
• Rule-based systems offer a structured approach to image analysis and world modeling.
• Representing world models within image/map databases is crucial for contextual understanding.

Purpose of the Study:

• To describe the organization of SPAM, a rule-based system for airport scene interpretation.
• To investigate the control of image processing and result interpretation using a world model.
• To present a method for representing world models in an image/map database.

Main Methods:

• Development of a rule-based system (SPAM) utilizing map and domain-specific knowledge.
• Integration of image processing control with world model interpretation.
• Representation of the world model within an image/map database.
• Analysis of high-resolution airport scenes with human and automated image segmentation.

Main Results:

• SPAM successfully labels image regions and groups them into consistent airport component interpretations.
• Interpretations are ranked based on spatial and structural consistency.
• Evaluations were conducted on three evolutionary versions of the SPAM system.

Conclusions:

• Rule-based systems can effectively interpret complex scenes like airports by integrating spatial and domain knowledge.
• The SPAM system demonstrates a viable approach to controlling image processing and aligning interpretations with a world model.
• Ranking interpretations by consistency provides a robust method for selecting the most plausible scene analysis.