1996/05/01 by Manfréd M. Fischer, Fischer, Manfred M., Sucharita Gopal +1
Computer Science · Earth and Planetary Sciences · Engineering · #Neural Networks and Applications #Remote Sensing and Land Use #Remote-Sensing Image Classification
paper · pdf · doi:10.57938/ff37896b-8bfa-4226-8c94-e2a398f3d07b
openalex publication_date 1996/05/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/01
Classification of terrain cover from satellite radar imagery represents an area of considerable <br/>current interest and research. Most satellite sensors used for land applications are of the imaging <br/>type. They record data in a variety of spectral channels and at a variety of ground resolutions. <br/>Spectral pattern recognition refers to classification procedures utilizing pixel-by-pixel spectral <br/>information as the basis for automated land cover classification. A number of methods have <br/>been developed in the past to classify pixels [resolution cells] from multispectral imagery to a <br/>priori given land cover categories. Their ability to provide land cover information with high <br/>classification accuracies is significant for work where accurate and reliable thematic information <br/>is needed. The current trend towards the use of more spectral bands on satellite instruments, <br/>such as visible and infrared imaging spectrometers, and finer pixel and grey level resolutions <br/>will offer more precise possibilities for accurate identification. But as the complexity of the data <br/>grows, so too does the need for more powerful tools to analyse them. <br/>It is the major objective of this study to analyse the capabilities and applicability of the neural <br/>pattern recognition system, called fuzzy ARTMAP, to generate high quality classifications of <br/>urban land cover using remotely sensed images. Fuzzy ARTMAP synthesizes fuzzy logic and <br/>Adaptive Resonance Theory (ART) by exploiting the formal similarity between the <br/>computations of fuzzy subsethood and the dynamics of category choice, search and learning. <br/>The paper describes design features, system dynamics and simulation algorithms of this <br/>learning system, which is trained and tested for classification (8 a priori given classes) of a <br/>multispectral image of a Landsat-5 Thematic Mapper scene (270 x 360 pixels) from the City of <br/>Vienna on a pixel-by-pixel basis. Fuzzy ARTMAP performance is compared with that of an <br/>error-based learning system based upon the multi-layer perceptron, and the Gaussian maximum <br/>likelihood classifier as conventional statistical benchmark on the same database. Both neural <br/>classifiers outperform the conventional classifier in terms of classification accuracy. Fuzzy <br/>ARTMAP leads to out-of-sample classification accuracies, very close to maximum <br/>performance, while the multi-layer perceptron - like the conventional classifier - shows <br/>difficulties to distinguish between some land use categories. (authors' abstract)