Abstract
This paper proposes a method of extracting landcover information of administrative districts in a prefecture by use of Landsat MSS data, and examines the applicability of its information to the statistical land-use surveys. The method is supported by the system which is composed of a personal computer and a large computer.
Errors of geometrical correction in the resampling of twelve combined Landsat MSS data are investigated. The land-cover information is extracted by a classification methdd on the basis of statistical verification. Fast Fourier Transform is applied in order to remove the effect of the clouds on the land-cover classification.
For the purpose of correcting the fluctuation of classification results, a method is proposed, which is based on the judgement of the change of land-cover pattern in the area divided into cell. The land-covers of classes "Paddy field", "Forest" and "Urban area" of the administrative districts are extracted from the Landsat MSS data in time-series and compared with the results of existing surveys.
The results show that the estimates derived from the Landsat MSS data approximately correspond to those of exisiting surveys as a whole, and suggest that the remote sensing technique by use of the Landsat is available to the land-use surveys in a large area.
