Abstract
Wetland vegetation classification was performed using casi image and measured detailed elevation data was overlaid with the casi image to analyze the relationship between the vegetation distribution and the slight elevation difference inside the wetland. This relationship exists because the water content distribution in wetland is related to the elevation difference. The analyses we have conducted are as follows. 1) An airborne spectral image (casi) with 2 m ground resolution was acquired over the Akai mire wetland on the June 2, 1994. When the image was acquired, the wetland vegetation was at the initial growth stage. In the analysis, radiance data of 3 channels (green, red, and near infrared band) which are known effective for detecting wetland vegetation difference were used. 2) 50 m grid digital elevation data were measure at both inside and surrounding wetland area by a total station. These data were interpolated and transformed into the 2 m raster data, to which the casi image was geometrically corrected and overlaid. 3 D view image of the casi data was produced using this digital elevation model. By comparing these images, relationships between wetland vegetation condition and the elevation difference were clarified. 3) k-means clustering (unsupervised learning) method was used to classify the casi image. 20 cluster classes were first calculated and merged into 8 vegetation community classes by vegetation investigations. We have succeeded in classifying even the sphagnum moss types. 4) The correspondence between the contour plot of detailed elevation and the wetland vegetation have shown the correspondence between the elevation differences and the vegetation types such as shhagnum moss types, grasses infesting on the sphagnum moss, and Pine trees at the edge of the wetland.
