Integration of multiple classic segmentation algorithms, oriented to the detection of edges, into a CommonKADS knowledge model

Abstract

The existence of specific domain and case problems when extracting relevant information from complex images have generated segmentation methods, as they are found in the bibliography, adapted to specific problems and situations. Working with satellite image of different areas, we developed multiple variations of a previous mobile windows algorithm from Cayula and Cornillon (1992). We started classifying our alternatives by their possible application, realizing the suitability of a general model of applicability. Our objective could be described as formalizing the knowledge about the application of multiple previous segmentation algorithms. This is a knowledge modeling task; deciding to use as knowledge system development framework, CommonKADS, which is, actually, a de facto standard in Europe. We present the first level TMD schema of the segmentation model (for edge detection algorithms) and the CML description of the general task and a specific sample method. Main conclusions include the interest of developing a knowledge model integrating existing segmentation algorithm in order to reuse them, and analyze their applicability in different domains.