Geocarto International · 1988 · 10 citations · 7 references
ComputingPhysical GeographyComputer VisionSpatial AnalysisSpatial Data AnalysisEdge DetectionFeature ExtractionImage SegmentationLand Cover ClassificationLand Cover Pattern AnalysisVegetation IndicesColor Edge DetectionColor Feature ExtractionForest RegionsLand Cover Feature ExtractionRegion SegmentationRegion-based SegmentationRemote Sensing Feature ExtractionUrban Feature Extraction
Abstract This paper presents a method for segmenting SPOT HRV data using color edge points to separate small regions with nearly constant color. Region segmentation using edge points has not been successful because edge detection operators yield the partially enclosed regions. The present method uses isograms of equal distances from the boundary points between the partially enclosed regions. The small enclosed regions (primitive regions) are separated, based on the local isogram maxima and their values. After SPOT data have been segmented, spectral and spatial features such as color, vegetation index, size and shape are extracted for each primitive region. It is shown that the primitive regions corresponding to typical land cover types (rice field, urban area, residential area, forest, highway and river) are successfully characterized by spectral and spatial features.
7