| Summary: | This article aims to present the change detection methodology as experienced in the use of optical remote sensing imagery (Landsat) and its pitfalls when multitemporal analyses are performed with pixel-based (raster algebra) techniques. The existing methodologyrecommends fundamental data preparation (geometric, radiometric, topographic corrections) and offers numerous change detection techniques. Regardless of the carefully performed preparationscertain noise remains, which can drastically weight the imagery comparisons. This noise behaves as a detected change and could have such a false effect on the identified change pattern (i.e. false, non-intrinsic changes) that the quantitative evaluation might fail.Since this noise originates from the pre-processing algorithms as well as the natural and technological conditions during data acquisition it can not be completely removed by data corrections. A multiresolution change detection approach is therefore proposed. Taking into account the neighbourhood and change information from joining different spatial scales, the multi-resolution approach effectively reduces the amount of false changes. In the discussion the remote sensing imagery for surface change detection is evaluated.
|
|---|
