Knowledge-Based Automatic Feature Extraction

Constructing geospatial databases is a tedious manual operation. Automatic 3-D feature extraction from 2-D images requires solving a number of problems. We present a plan to attack this task using a combination of tools: reconstruction and reasoning in 3-D, use of multiple sources of data, perceptual grouping, use of context and domain knowledge, use of previous maps and models, and limited use of human input where applicable. 1 Objectives Geospatial databases are important for a number of battlefield awareness tasks such as mission planning, mission rehearsal, tactical training and damage assessment. Other applications include intelligence analysis for site monitoring and change detection. Geospatial database requirements may vary for the different tasks, but generally knowledge of terrain elevation, surface features and cultural features is needed. Our task focuses on the automatic detection and description of cultural features, particularly buildings. The importance of cultural fe..