A STRUCTURAL APPROACH TOWARDS DRAINAGE PATTERN RECOGNITION (REMOTE SENSING, PHOTOINTERPRETATION, IMAGE UNDERSTANDING, LANDFORMS)

The drainage pattern is the most important single characteristic for determining the structure and type of a landform. This research effort has developed a structural pattern recognition approach and the associated software for the identification of drainage patterns. A structural model for drainage pattern description was developed. This model included the decomposition of drainage patterns in a hierarchical organization of line segments. Every line segment was characterized by a set of attributes. The arcs connecting the nodes of the hierarchy were characterized by attributes of the relations between the line segments of the connected nodes. Drainage pattern attributes were designed to express the topologic and geometric relationships among the high-level segments of the drainage patterns. These attributes were derived as composite attributes or attribute functions. Classification was performed by a decision tree that facilitated the testing of the occurrence of certain key attributes between an input pattern and a stored description of a prototype. The DPA system has been evaluated with real and artificial patterns from all the eight classes and it has performed satisfactorily.