Clustering Algorithm in ILWIS GIS for Classification of Landsat TM Scenes (Mecsek Hills Region, Hungary)

Polina Lemenkova (1) , István Elek (2)

1 OUC - Ocean University of China
2 ELTE - Eötvös Loránd University

Current research has been performed at Eötvös Loránd University, Institute of Cartography. - RESEARCH EMPHASIS: application of clustering spatial analysis of the open source ILWIS GIS. - RESEARCH AIM: agricultural mapping of land cover types: south-west Hungary, Mecsek Hills. - RESEARCH PROCESS: Landsat TM scenes were classified into different land use types: natural veg- etation coverage, anthropogenic areas and agricul- tural fields, sub-divided to various crop types. - RESEARCH OUTPUT: three independent agricultural thematic maps of land cover types for years 1992, 1999 and 2006, created in ILWIS GIS. - RESEARCH DATA: Landsat TM on 14.09.1992, a) 09.08.1999 b) 19.07.2006 - METHODS: The research methodology is based on cluster classification algorithm available in ILWIS GIS. The work is organized in several research steps summarized in the research workflow: The research area was classified into a set of land cover categories, labelled to following land units: 1) winter wheat 2) spring barley, 3) maize 4) sugar beet 5) maize for ensilage 6) oak and beech forests 7) potato 8) other crops 9) shrubland 10) water 11) not agricult. areas 12) grassland 13) other land cover types Field crops (e.g. maize, winter wheat) were detected on the images. The species with unclear nature of crop or not easily recognized were defined as ’other crops’. A Google Earth aerial imagery was used for visual control inspection. Once all clusters are grouped, the layout was created using representation palette defined in the domain ’Land Cover Types. The research results in 3 maps of land cover types for 1992, 1999 and 2006. Clustering method can be applied for other agricultural areas, since it enables objective classification in regions with high land heterogeneity and complex landscape structure. PRINCIPLE Clustering is based on principle of spectral distinguishability of digital cells. The Digital Numbers (DNs) of pixels create unique spectral signatures for various objects. Clustering extracts info on pixels and analyses similarity of their DNs. Pixels with similar value of DNs are assigned to thematic categories (clusters). The grouping is done according to pixels’ similarity, which is larger within a group than among other groups. ADVANTAGES Clustering is objective technique, useful in situations when fieldwork is not available. It enables to avoid misclassified pixels and ignore external factors (e.g. atmospheric conditions), which significantly facilitates spatial analysis of the images.

Mots clés: Image processing Digital image processing; Satellite image interpretation; Landsat Imagery; Landsat TM; GIS & Spatial Analyses; GIS Data Modelling; GIS Analysis; Cartographic Generalisation; Cartographic analysis; Land cover and land use; Land cover and land use changes

Domaines: Vision par ordinateur et reconnaissance de formes [cs.CV]; Traitement des images [eess.IV]; Synthèse d'image et réalité virtuelle [cs.GR]; Ordinateur et société [cs.CY]; Interface homme-machine [cs.HC]; Traitement du signal et de l'image [eess.SP]

Polina Lemenkova, István Elek. Clustering Algorithm in ILWIS GIS for Classification of Landsat TM Scenes (Mecsek Hills Region, Hungary). 3rd International Conference ’Geosciences and Environment’. Near-Surface Geophysics, May 2012, Belgrade, Serbia. Geosciences and Environment The 3rd International Professional Conference, 2012, ⟨10.13140/RG.2.2.36448.74244⟩. ⟨hal-01972872⟩