A machine learning approach to automated construction of knowledge bases for expert systems for remote sensing image analysis with GIS data
- Univ. of South Carolina, Columbia, SC (United States). Dept. of Geography
- Westinghouse Savannah River Co., Aiken, SC (United States)
Knowledge-based remote sensing image analysis with GIS data is acknowledged as a promising technique. However, the difficulty in knowledge acquisition, a well-known bottleneck in building knowledge-based systems, impedes the adoption of this technique. Automating knowledge acquisition is therefore in demand. This paper presents a machine learning approach to automated construction of knowledge bases for image analysis expert systems integrating remotely sensed and GIS data. The methodology applied in the study is based on inductive learning techniques in machine learning, a subarea of artificial intelligence. It involves training with examples from remote sensing and GIS data, learning using the inductive principles, decision tree generating, rule generating from the decision tree, and knowledge base building for an image analysis expert system. This method was used to construct a knowledge base for wetland classification of Par Pond on the Savannah River Site, SC, using SPOT image data and GIS data. The preliminary results show that this method can provide an effective approach to integration of remotely sensed and GIS data in geographic information processing.
- Research Organization:
- Westinghouse Savannah River Co., Aiken, SC (United States)
- Sponsoring Organization:
- USDOE, Washington, DC (United States)
- DOE Contract Number:
- AC09-89SR18035
- OSTI ID:
- 238540
- Report Number(s):
- WSRC-MS--95-0433; CONF-9604133--1; ON: DE96009650
- Country of Publication:
- United States
- Language:
- English
Similar Records
Comparison of two inductive learning methods: A case study in failed fuel identification
Comparison of two inductive learning methods: A case study in failed fuel identification