Automatic feature extraction techniques were developed for use with digital images and map data to assess the feasibility of employing expert systems for map revision. Urban test areas were selected and SPOT images, map separates, and printed map sheets acquired. The digitized map and image data were placed in register to create a cartographic database suitable for use with a prototype expert system optimized for the extraction of building features.
Input images were segmented with the region growing method using optimum threshold values derived from map data. Twenty descriptors of shape, size, and tone such as area and elongatedness were calculated for each of the segmented regions using these descriptors. The expert system was also designed to direct the image processing routines with specific instructions (“how to analyze”) applied to focused areas (“where to look”) in the iteration process.
The map data were useful for determining initial parameter values for image processing and for change detection of existing features. An expert system approach permitted control of the iterations required for feature extraction and the refinement of threshold values. The accuracy of feature extraction increased as the image pixel resolution was improved. In order to realize feature extraction results comparable to those achieved by human interpreters, digital images must be resampled to pixel resolutions of one-half to one-fourth the original pixel dimension.