Image Processing and Analysis of Change Detection in the Land Cover Types of the Sudetes by Idrisi GIS

The presentation reports testing Idrisi GIS methods of raster images processing using Idrisi GIS software. Supervised and unsupervised methods of image classification are tested and presented: Maximal Likelihood (MAXLIKE) algorithm and Minimal Distance (MINDIST) algorithm. The raster images cover period of 20 years 1984, 11 July, 1992 and 2003, 17 August. MAXLIKE approach evaluates the standard deviation of the reflectance values above the mean. The slowest technique but more accurate classification (provided the training sites are good). MINDIST approach is the simplest and fastest method of all classifiers. However, prone to incorrect classifications. The results show changes in the land cover types since 1984 in the test area of Sudetes, Poland. Current work reports detected changes in land cover types using Idrisi GIS and technical questions of satellite image analysis and processing.