Takashi Hoshi

A remote sensing data classification method using self-organizing map

A supervised classification method using a self-organizing map (SOM) is proposed to classify remote sensing data. The SOM structure is composed of two layers. One is an input layer with nodes corresponding to spectral bands. The other is an output layer with square array of nodes. First, a feature map on the output layer is generated by inputting training data into SOM. Each node in the feature map cannot be corresponding to a category because the number of nodes is generally greater than those of training data. Thus, a cluster map is generated by comparing differentials among weight vectors in nodes. Secondly, the training data is re-inputted into the cluster map to find the relationship between clusters and categories, that is, the cluster including a fired node is labeled as the category to which the training data belongs. In consequence of mapping, the category map is obtained from the feature map. The proposed classification method extracts liquefied area in Kobe (Japan) damaged by the 1995 Hyogoken Nanbu earthquake using the SPOT HRV data and the category map. As an experimental result, it is shown that classification accuracies of the proposed method are higher than those of the maximum likelihood and the backpropagation methods.

Polarimetric SAR data classification method using the self-organizing map

In this paper, we introduce a supervised classification method, which differentiates polarimetric SAR data into three categories using a self-organizing map (SOM) and a counter propagation learning approach after identifying the appropriate scattering classes. This classifier produces category maps corresponding to the Kohonen layers using training data for each scattering class. The SAR data are classified by inputting both like- and cross-polarization power elements into the learned SOM. In the experiment, PI-SAR data are employed since the resolution of aerial SAR data is higher than that of SAR data obtained from space. The proposed method yields higher-accuracy classifications than do conventional methods.

Application of equivocation quantification to fuzzy data

Considers multispectral remote sensing images where every pixel in the data corresponds to a mixed pattern on the ground. In the image processing it is necessary to consider one pixel as one pattern on the ground in one case, and one pixel as multi patterns in another case. The former is used to calculate the classification scores in order to assess classification results. Since the classification score is calculated based on crisp data, the original equivocation quantification can be used to assess the classification result. However, since the later one is fuzzy data, i.e., the membership function value is a real number from zero to one, the original equivocation quantification cannot be used to assess the classification result directly. The authors introduce a new concept called pseudo-probability to adapt fuzzy data to calculating the equivocation quantification.<<ETX>>

Formating of the Parameter Blocks in the CCT of GMS“Himawari”

GMSのCCTデータのパラメータ部と画像の一部のフォーマット変換とリスト処理の手法を開発した。このフォーマット変換は, データコントロールパラメータと記述文コントロールパラメータを用いて, バッファメモリ上で編集しながら処理を行うものである。このフォーマット変換を用いれば, 見出し, ページ, 項目番号, 記述文, データの位置, データの長さ, 副記述文を付加し文字型に変換されたパラメータ部のデータが得られる。また, 画像データ部の1ラインのダンプリストが10進数と16進数で得られる。この手法は, LANDSAT-MSS, LANDSAT-RBV, NOAAのCCTデータにも応用できる。LANDSAT-RBVとNOAAに関しては, すでにフォーマット変換を実行している。この変換手法を利用することによる利点はつぎのとおりである。1) パラメータ部のデータの内容が理解しやすくなる。2) 画像データのダンプリストにより, 画像のレベルを解析以前に知ることができる。3) 将来確立が予想される, リモートセンシングデータの統一的なデータ管理システムに応用できる。

Pattern Recognition by the Colour Tone of Land Use Map

In order to recognize the parrerns represented in a map, picking out of distribution patterens in the map has to be done as pre-processing. Quantitization method was used to pick up the distribution patterns and B. G. R. and B. W. methods are applied for quantitization. Picking out of colour tone distributions was carried out by applying this method. The results show that the accuracy of this method is farily good. But we, also, find some misrecognitions and luck of similitude of patterns between original and obtained figures. In this paper, so, improvement of accuracy are attempted by changing the scale of colour slide obtained from the original map and by increasing nomber of sampling points. As another processing technique, continuous tone negative film separeted each colour tone in the original map is investigated.As these results, it is clear that distribution patterns represented by several colour tones are able to pick up fairly well by B. G. R. and B. W. method, and they will also be picked out simply by using B. G. R. method only.

Pattern recognition of a few colour tone by Continuous tone negative

Looking back upon the history of photogrammetry, in short, it seems that in quantitative photogrammetry, the weight had been poured into the determination of the image position.But in recent years, quantitative use of digitalized information of pictures has developped more rapidly when the validity of information recorded by photo was recognized more, and it seems which is developping to wide shere as characteristic information of the image position from narrow sphere as determination of the image position.Recently, this author has thought that quantitative information of the image position is useful in many fields as civil engineering, medical science, agriculture, and physical sciences, and has executed to densito-measurement of the X-ray photograph, photo of soot and smoke of factories, lights and shades on map and others.However, it is hard that pattern recognition of the pictures are expressed only by lights and shades element. This paper express pattern recognition of a few colour tone.