高分解能航空機MSSデータを用いた自動分類による市街地におけるみどりの抽出

抄録

The aspects of urban vegetations are important information for urban environmental assessment. The information, however, should be updated to accomodate the changing conditions. For this purpose, this paper discusses about an extraction method of urban vegetations by means of supervised classification using airborne Multi-spectral Scanner data. An effective method is proposed. A pixel by pixel maximum likelihood method was used for supervised classification. CC-14, one of 14 channel combinations got the highest classification accuracy. It consists of 3 visible bands, 2 near infrared bands, and 1 thermal infrared band. This method was applied to the MSS data which had been observed with high spatial resolution of 1.2 m×1.2 m in residential areas in the daytime of a clear summer day. Classification results were compared with ground truth data at the view points of distribution aspects of vegetations and vegetation ratio to the land. It is clarified that the extraction method using CC-14 gave good agreement with the ground truth resuits and can extract vegetations whose area is more than an area of 6 m^2.