Gis-driven analyses of remotely sensed data for quality assessment of existing land cover classification

Automatization of processes for revision and updating existing GIS information is essential for the modern maintenance of spatial databases. The integration of remotely sensed multi-spectral data into the process of database revision is affected here by the implementation of GIS-driven analyses. The adoption of the GIS-driven principles, provide also an accurate geographical basis for a future supervised classification of the spectral data. The goal of the present research was to define and develop an automatic quality assessment method for the Land Cover classification layer of the Israeli National GIS database. During the experiments on multi-spectral remotely sensed data, effort was carried out in attempt to define "typical" spectral ranges as statistical maximum-likelihood criteria for the classification of each of the land cover phenomenon. These ranges were envisaged to characterize each of the land cover classification groups and to provide quantitative criteria for the definition of various groups of land cover type-classes. The definition of a typical-spectral-variance was executed on the basis of visual, multi-spectral and index bands of remotely sensed data. The decision whether existing GIS classification match the new image reality was made by statistical criteria of maximum likelihood for each investigated land cover type, according to the results of each and every spectral band. The study was based on multi-spectral data of the CASI airborne scanner and the space borne IKONOS data. Acquisition of the various types of information (spectral and spatial) was done according to the GIS-Driven approach of spectral data analyses that also permit the treatment of any spectral phenomena in a local coordinate system. The proposed method was developed on the basis of a study area (~50 sq. km.) and was tested on a larger control area. Image reality and field verification (both by land and air) proved the method of GIS-Driven quality assessment to be a promising solution for a revising process of large core spatial Data Bases.