| 概要 |
GIS上でデジタルオルソフォトに対して画像分類の手法を用い,一定期間にわたり裸地化したまま放置されている不成績造林地を抽出する方法について検討した.まず同一地点において,時点の異なる2枚のデジタルオルソフォトを用意し,それぞれについてGIS上で画像分類を行った.画像分類に置いてはオルソフォト画像に加えて,テクスチャ画像を用い精度の良い分類結果を得た.次に両時点において裸地であると分類された画素を抽...出し,これを一定期間にわたり裸地化したまま放置されている林地であると定義した.この画像と小班界ベクターデータを重ね合わせ,造林地の中で裸地化したまま放置されている林地を抽出し,これを不成績造林地とした.さらに小班界ベクターデータを利用して2時点重ね合わせ分類画像を集計し,人工造林された各小班のなかで不成績造林地に判定された画素の割合を求めた.最後に対象小班が不成績であるかどうかの判定基準として,小班内での不成績造林セルの割合が35%以上であるという基準を用いると,その判定結果は現地調査から得られるものに近づき,本研究で採った手法により不成績造林地の抽出を高い精度で行うことが出来ることが分かった.ただし今回の手法では,造林木の成長が悪く代わりに広葉樹が進入して繁茂しているといった種類の不成績造林地は抽出することができず,本方法に不成績造林地抽出の限度があることも明らかとなった.
Many young planted sugi (Cryptomeria japonica D.Don) and hinoki (Chamacyparis obtusa Endl.) trees in mountainous regions of Kyushu have been removed by deer's (Cervus nippon nippon Temminck) herbivory. Stands denuded by deer's herbivory have been endangered to the point of collapse. The forest status must therefore be monitored. Orthophoto graphs have map-like geometry and useful as accurate image bases for geospatial analysis, so we proposed a monitoring technique with digital orthophotographs. To monitor the forest status, we used an image classification method base on GIS. To increase the accuracy of the classification, we applied the following three steps. First, we calculated texture images as the standard deviation of a cell value within the neighboring nine cells and added these as input layers. Through this process, we classified digital orthophotographs into three classes, broadleaf forest, coniferous plantation and moorland. Second, we classified two-temporal digital orthophotographs and compared them to detect changes caused by forest operations. After this process, we extracted cells that were classified as morland in both images as unchanged moorland. Third, using the GIS data, which has subcompartment boundary information and forest register information, we extracted cells including sugi or hinoki plantation subcompartments. Finally, we calculated the percentage of area classified as unchanged moorland in each sugi or hinoki plantation subcompartment,. We applied these processes to part of the Kyushu University forest in Miyazaki where natural forests had been cut and sugi or hinoki planted after 1978. If the percentage of area classified as unchanged moorland exceeded 35% for each sugi or hinoki plantation subcompartment, we defined these subcompartments as unsuccessful plantation. Judgements based on this rule agreed well with reconnaissance. However, these processes couldn't define unsuccesful plantations where invasive broadleaf trees had grown insted of sugi or hinoki, so we found this procedure was limited in use.続きを見る
|
Mendeley出力
