ミクセルの画素内緑被配置推定による都市緑被分布図の精細化

概要

論文の詳細を見る
Detailed maps of vegetative cover are used for analyzing and evaluating urban environments. The use of remote sensing data is an effective way to create such maps because these data are obtained with a large area repeatedly. However, since much small vegetation exists in urban areas and such vegetation is observed as mixels, it is difficult to classify them into vegetative areas and non-vegetative ones. This paper utilizes a neural network technique to estimate the location of vegetation in the mixels from the image which is generated from the occupation ratio of vegetation in a pixel. The scope of this estimation method is when the spatial resolution of the data is smaller than vegetation. In this case, the mixels exist only in the boundary and parts of the vegetation constituting mixels are contained in the neighboring pixels. Therefore, this study assumes that the location of the vegetation in a mixel can be estimated from the occupation ratio of the neighborhood pixels. Several patterns of vegetation cover locations in the mixels were estimated from eight neighborhood pixel patterns using the neural network technique. As a result, the location of the vegetation in the mixels was able to be estimated when mixels were divided into nine pixels, and the effectiveness of this method was demonstrated from detailed vegetation shapes.
社団法人日本建築学会の論文
2001-05-30