2005 年 60 巻 6 号 p. 1207-1210
In previous research a singular value decomposing method with some restraint conditions was used to estimate unknown endmembers and coverage in agricultural land. However, it is limited to the ideal case I which observed spectra don’t include noise. Hence it is necessary to verify the feasibility of this method to spectra obtained from agricultural area.
In agricultural area, two cases are assumed: observed spectra have a large and a small deviation. If observed spectra have a large deviation, it is more difficult to estimate endmembers. However estimating endmembers is possible using the singular value decomposing method even if observed spectra have a large one. Consequently, it is showed that this method is applicable to even real agricultural land.
Additionally, a new index of plant characteristic was proposed. NDVI (Normalized Difference Vegetation Index) has used in evaluating plant characteristic as usual, however it is affected by coverage significantly. Some researches have reported that NDVI expresses the plant characteristic when the influence of coverage can be neglected. In this study, the influence of coverage was removed to express degree of plant activity numerically.