The split window (SW) method is a technique for estimating surface temperature by the linear combination of the brightness temperatures measured from a satellite in two considered spectral channels in the thermal infrared window. The multi-channel (MC) method is an extension of the SW method; it uses the brightness temperatures measured in two or more channels for the estimation. The extended multi-channel (EMC) method is an extension of the MC method; its objective variable is not surface temperature, but the upward brightness temperature at surface level in each channel.
In the case of applying these methods to land observations, the error based on the uncertainty of surface emissivity should be considered. We, therefore, researched the relationship between the error and the uncertainty of emissivity using the simulation data for NOAA12 AVHRR (channel 4 and 5) and EOS-AM1 ASTER (channel 10 to 14) under global conditions. The main results are as follows:
1) The error of the SW method increases linearly with the uncertainty of emissivity. This result validates the Becker's error formula.
2) The error based on the uncertainty of emissivity is significantly reduced by optimizing the equation to the uncertainty of emissivity using various emissivity data. And it is also reduced by increasing the number of channels and determining the equation with the emissivity data from spectral library.
3) The error of the EMC method based on the uncertainty of emissivity depends on the channel, and converges as the uncertainty of emissivity increases.
4) In no consideration of emissivity, the RMS error of the MC method is 0.8K for both AVHRR and ASTER. In the case that the domain of emissivity is from 0.98 to unity, the error is 1.0K for AVHRR and 0.7K for
ASTER. And in the case that the domain of emissivity is from 0.95 to unity, the error is 1.5K for AVHRR and 1.0K for ASTER. For this domain, the RMS error of the EMC method for each channel is from 1.2 to 1.5K for AVHRR and from 0.9 to 1.1K for ASTER.

抄録全体を表示

本研究は,衛星マルチスペクトルデータから得られる画像特徴の判読支援を目的として,遺伝的アルゴリズムを導入した画像特徴・強調処理手法を開発し,その適用性について検討したものである。コントラスト,エッジ情報等の画像特徴は,土地被覆状態,線構造,地形形状等を判読し,分析する際の支援情報として多用されている。しかし,処理後の画像(以下,特徴画像)では,元画像に比べて情報量が減少するために,特徴画像をカラー映像表示装置上に表示する際の強調処理(コントラストストレッチ処理等)効果が十分に得られない場合がある。そこで,本研究では,遺伝的アルゴリズムを用いて画像全体のエントロピー(適応度関数)を増加させ,特徴画像そのものの情報量を増加させるとともに,画質を向上させる方法を提案した。本研究の成果は,次の3点に要約できる。
1)特徴画像のエントロピーを増加させる上で,本研究で設計したアルゴリズムに基づく遺伝的操作は有効に機能することが判った。
2)提案手法を適用した処理画像は,従来の特徴画像に比べて「画質」および「情報量」の両面において向上しており,画像解釈における支援情報として有用性の高いことが確認された。
3)さらに,提案手法は,コントラスト画像,標準偏差画像,エッジ強調画像といった種々の特徴画像に対する適用効果が確認され,汎用化への指針も得られた。

抄録全体を表示

The objective of this study is to construct the evaluation algorithm for the land cover change detection using satellite multispectral data. The post-classification comparisons based on the several supervised or unsupervised classification methods are the most commonly used methods of quantitative change detection. It requires a complete classification of the individual dates of the satellite data. Unfortunately, every error in the individual data classification map will also be present in the final change detection map. Therefore, it is imperative that the individual classification maps used the post-classification change detection method be as accurate as possible. It should be recognized that there are still many difficulties on the satellite-based change detection of the land cover. So, we divided the research approach of extracting and analyzing the land cover change into two stages; one is the evaluation on the changed "area", and the other is the changed "class" of the land cover. There are some limitations to evaluate the changed "class" accurately with the image processing based-method. In this study, we focus only on the evaluation of the changed "area" of the land cover. The proposed algorithms for the land cover change detection are as follows;
Step-1) Preparation of the two kinds of satellite data observed at the different time. Let's say those "Data-A" and "Data-B", respectively.
Step-2) As the preprocessing for the Data-A and the Data-B, the geometric correction and the radiometric normalization were executed.
Step-3) Through the 3 x 3 window operator, the dissimilarity measure between Data-A and Data-B is calculated, for the TM band-2, band-3 and band-4, respectively. Then the "dissimilarity images" are produced in each band.
Step-4) The Land cover Change Detection map (termed LCD map) is produced by assigning the similarity images to the red, green, and blue image planes, respectively.
Step-5) Interpretation of the land cover change detection map. In comparison with the change detection results of the post-classification comparison method based on the maximum likelihood classification, we conclude;
1) The LCD map could be made through the simple procedure, that is important factor as the image processing and analysis method on the land cover change detection.
2) The LCD map could highlight the land cover changed "area" observed not only in all band, but also in each band.
3) Based on the color composite information of the LCD map, we can easily interpret the changed situation of the land cover, such as the paddy field, the farmland, the grass and forest area, etc. The LCD map is useful as a "supporting information" to extract and analyze the land cover changed area, as well as for the detail field investigation. The proposed algorithms for making the LCD map and its interpretation should contribute to the feature research activities on the land cover change detection.

抄録全体を表示

The eruption clouds by the strong fissure eruptions on November 21, 1986 at Izu-Oshima volcano, Izu Islands, Japan are well detected in GMS imageries. The top altitude of eruption clouds estimated from the coldest surface temperature of GMS infrared data is 7-9 km, but this value is distinctly lower than the reliable heights of 10-16 km determined by the observations from the ground and weather radars. The northern edge of the extent of eruption cloud in GMS imagery is apparently over the southern tip of the Boso Peninsula. These two apparent observation results can be corrected with considerations of the effect of the GMS's parallax angle. The reason of the underestimation of the top altitude may be due to heat discharges through the surface of eruption cloud from internal heat source during the development of eruption clouds.

抄録全体を表示