本多 嘉明

自動分類の問題には, コンピューターが理解できるような植生のカテゴリを記述する手法やそのカテゴリにしたがって分類を行うことなどがある。分類手法には, 教師付き分類法と教師無し分類法の二つがある。これらの手法には自動的に植生のスペクトル特性とカテゴリを結びつける手法がない。著者らはGASC手法をべースとする画像不変式を見出した。<BR>本研究ではその不変式を用いて, ベトナム南部 (Ninh Thuan, Binh Thuan and Lam Dong provinces) においてGLIシミュレーションデータを用いて研究を行った。その結果, 処理時間を短くすることができた。また, この手法で高精度に分類を行うためには広範囲のグランドトゥルースデータが必要であることがあきらかとなった。

A new composite method is proposed. This method is designed to apply to the global data set and it is composed of three consecutive criteria which are based on NDVI, brightness temperature and scan angle, respectively. This method is applied to the global data set with 4km resolution which is processed by Center for Environmental Remote Sensing (CEReS), Chiba University. The comparison with Maximum Value Composite (MVC) and maximum brightness temperature composite showed some advantages of this method. Clouds over low NDVI region as desert are eliminated due to brightness temperature criterion. Neighboring pixels are more frequently selected with same data due to scan angle criterion. It shows that the atmospheric condition and the observation geometry is similar over a wide area. The scan angle distribution is centered on nadir, therefore, effects of atmosphere and bidirectional reflectance, size of footprint and overlap between adjacent pixels are reduced.

The occurrence of a great forest fire in Indonesia in this summer gave impacts on various phases in social and economic activities. The smog caused by the fire influences neighboring countries in addition to Indonesia. To prevent the damage area extending, we need to get information about present conditions on the disaster. However, since a target area is broad in this case, it's difficult to get needed information.<BR>We have used Defense Meteorological Satellite Program (DMSP) and Geostationary Meteorological Satellite (GMS) to monitor this disaster. DMSP detects slight emission such as fires at night and GMS collects data on cloud every hour. By using two different types of satellite data, we can get information on the location of hot spots and spread of smog caused by the fire. Time sequential data we have processed from GMS clarifies movement of the smog. From DMSP, we can detect the locations of hot spots.<BR>On September 21, there were many hot spots in the southern part of the Kalimantan Island. On September 30, however, a number of the hot spots in that area became smaller. On the other hand, a number of hot spots in the Sumatra Island didn't change for more than 10 days. Smog generated in the Kalimantan Island moves westward; Singapore, Malaysia, and so on. The spread of smog still remains as of September 30.<BR>As mentioned, by using two satellites of different characteristic, we can clarify the change in conditions of the hot snots and the movement of smoe venerated by the fire.