Automatic Land Use/Land Cover classification system with rules based both on objects attributes and landscape indicator

Recently, there is an increasing demand for information on actual land use/land cover (LU/LC) from planning, administration and scientific institutions. Remote sensing combined with GIS tools can give quick reply providing timely information products in different geometric and thematic scales. Anyway the effort to make the land use map by visual interpretation is still very high and cannot keep up with the development pace. On the other side automatic procedures do not assure to follow detailed and well- structured land use nomenclature if it is not performed by a customized learnig system. This new approach is required to incorporate automated image classification to human image understanding. In this context the here proposed T-MAP application combines segmentation tools with an hybrid classification technique (a rather new trend in image classification) and a rule-based thematic categorization depending on information both at pixel and object level. Its rule-based system lets the user define thematic assignments by building rules based on feature attributes (i.e. membership cover class percentage, confusion index, etc.) and on landscape analysis (spatial metrics and pattern indicators), taking, in this way, also advantage of the human land use understanding. The output is a thematic map characterized by a custom-designed legend and a reasonable performance in terms of accuracy, number of extractable classes and legend detail.