抄録
Remote sensing has been recognized as one of the most power full tools in the collection of geographical data. However, the application of remote sensing technology is rather underdeveloped compared with that of hardwares such as LANDSAT (ERTS) or scanners which show rapid progress in recent years. Therefore, the major objective of the research work done by present author is to develop the operational analitytical systems for fast image interpretation.
This paper deals with digital remote sensed image processing in land use and vege-tation. An analysis was made on discriminant analysis of land use classification using ERTS multi-spectral data. As the result, in 13 class discrimination, the rate of correct correspondence to the ground truth was about 40%. In 5 class discrimination, the rate was increased up to 69%. The main causes of errors are caused by 1) the scale affection of the size of picture cell, 2) lack in appropriateness in land use classification, and 3) noises in images.
Another analysis was made on classification of vegetation using digitalized multi-spectral photographs cdata. Interactive image processing system with color display unit was develo-ped for the analysis. This digital color system with mini-computer can be used not only for additive color display but also for the display of numerically processed images or for the pattern extraction by multidimensional threshold method. Using this system, vegetation was classified into the categories, and each categoriy is displayed in different colors.
As for the future prospects, further geographical applications are strongly recommended for the purpose of accumulating basic knowledge. The technological development, such as the introduction of pattern recognition methods in spatial context, is also indispen-sable for the future analysis.
