Estimation of the rice-planted acreage is divided into two processes, discrimination of the rice-planted area and totalization of the riceplanted area. Satellite radar remote sensing is effective for discriminating the rice-planted area, because the growing season for rice in Japan, overlaps with the rainy season. Furthermore, the rice-planted area can be accurately totalized by referring to agricultural plots vector data. In order to improve the estimation accuracy of the rice-planted acreage, a method which discriminates the rice-planted area using satellite radar data and totalized the rice-planted area referring to agricultural plots vector data was proposed in this paper. The proposed method was applied to RADARSAT data to estimate the riceplanted acreage for a test site in Koshiji, Niigata prefecture. The estimated rice-planted acreage obtained by the above method was compared with that acquired by QuickBird data. As a result of accuracy verification, the accuracy of estimation using the above method was 85%.

抄録全体を表示

This paper proposes a new method of LIDAR data processing to automatically extract building and 3-D city modeling. This method exploits image processing technique including Hough transformation. The process of building extraction starts with DSM segmentation according to elevation values. Among these fragments, extracting process limits candidates of building fragments to those that have elevation higher than a specified threshold value and the DEM at the same location. Hough transformation is executed to boundary pixels of each candidate of building fragment and then fit to linear segments. Edges in corresponding parts of aerial images can also be used to refine boundaries of buildings. Each fragment is polygonized with these line segments. After the extraction, 3-D-building model is created with these polygons such that each polygon has vertical wall from the top of building to the ground (DEM) . Texture images are pasted onto each building, including walls. The most appropriate texture image is selected from aerial images according to geometry between building faces and external parameters of the photos. The method has been tested with LIDAR data of Shinjuku and Ginza area in Tokyo.

抄録全体を表示

It is very important to control the water stress at the growth stage of Satsuma Mandarin for high sugar content. It is known that a shift of the red edge wavelength of spectral reflectance to shorter wavelengths is caused in water stressed leaves. The authors investigated the transition of the water potential and the red edge wavelength of spectral reflectance in the actual fruit cultivation. Its period was from July to November in 2003. Consequently, the transition of the red edge wavelength was the same as the transition of the water potential in filmmulching on the ground surface of the fruit trees. Tow main results were obtained. First, an adequate correlation between the averaged sugar content in harvest fruit and amount of change of the red edge wavelength at the growth stage of fruit was indicated from the results of measurements. Second, it is assumed that the individual differences of mandarin trees, such as the amount of fruit, volume of tree, and nutritive condition, influence the red edge wavelength.

抄録全体を表示

This paper focuses on the evaluation of compositing algorithms of MODerate resolution Imaging Spectroradiometer (MODIS) . The atmospheric effects of channel 1-7 have been investigated by radiative transfer simulation using 6S code over variety of aerosol and climatic conditions. As a result, blue wavelength (channel 3) was found to be the most subject to atmospheric effects and its apparent reflectance values were higher than other channels. However, this method is subject to select cloud shadows as a optimum values. To avoid this problem, the combination of minimum blue constrained with thermal channel criteria is newly developed based on the assumption that the cloud shadow is cooler than normal pixels. These two approaches, to say, the minimum blue criterion (MinB) and the minimum blue and maximum thermal criteria (TMinB) have been tested along with five approaches including the maximum NDVI criterion (MaxN), the maximum temperature criterion (MaxT), the NDVI criteria followed by minimum scan angle (NMinS), the thermal criteria followed by minimum scan angle (TMinS) and the NDVI and thermal criteria followed by minimum scan angle method (NTMinS), to improve on the heritage of Advanced Very High Resolution Radiometer (AVHRR) . Then the 10 day composite images over Japan main islands have been generated using the above mentioned seven approaches. The comparative performances of these seven methods are judged from five criterion including cloud screening, satellite zenith angle, smoothness of the image, snow/ice identification and cloud shadow. It is concluded that the TMinB method produces fairly consistent results with possible refinements and more sophisticated methods to escape from cloud shadows.

抄録全体を表示

We have examined positioning availability and accuracy of OmniSTAR, a wide area DGPS service, and StarFire, a global DGPS service, comparing with RTK and the DGPS service by Japan Cost Guard (JCG) . The experiments were for utilization of GPS in urban area. There were four experiments, i.e. fixing antennas on known points, carrying antennas by hands and brief stop on known points, revolving antennas, and carrying antennas by a vehicle. Accuracy of StarFire and OmniSTAR are not higher than the JCD's DGPS, large errors were sometimes observed for StarFire. Availability of DGPS is much higher than RTK. But availability of OmniSTAR was sometimes decreased by microwave wireless facilities.

抄録全体を表示

教師なしニューラルネットの一種である, 自己組織化マッピングの学習を通じて密度マップが生成できる。この密度マップはクラスタ境界を反映していると考えられる。ここで提案する自己組織化マップに基づく画像クラスタリング手法は, 特徴空間におけるクラスタ分布が複雑に入り組んでいるような場合にも十分なクラスタリング性能を達成し得るように考案したものである。学習過程を経て得られる密度マップから複雑に入り組んだクラスタ境界を推定し, これに基づきクラスタリングするものである。このような特徴空間における複雑なクラスタ分布に対応し得るクラスタリング方法として, 分離統合法によって対処することも可能であるが, この場合はパラメトリックであり, 対処療法的である。従って, すべての場合に適用できる保証がない。本提案手法は, 一方, すべての場合に対処し得るが, 複雑になればなるほど収束に時間が係り, 局所解に陥ることもある。また, 既存のクラスタリング方法に比し, 最終クラスタ数を限定することを前提としていない等の相違がある。
シミュレーション画像及びランドサットTM画像を用いて, 提案手法を最尤法分類, k平均法とクラスタ性能を比較することにより評価したところ, 提案方法の優位性を確認した。シミュレーション実験ではクラスタ内分散の小さい場合は収束が速いが, 分散が大きく複雑な分布の場合は収束が遅く, 局所解に陥ることもあることが分かった。また, TM画像分類実験では, 代表的な教師付き画像分類法である最尤法による分類結果に近いクラスタ結果を確認した。さらに, クラスタ問分離度がk平均法に比し, 15.9%向上することも分かった。

抄録全体を表示

既存の階層的, 非階層的なクラスタリング方法は, 基本的に線形分離が可能なクラスタ分布に対してはよくクラスタ化できるが, それが不可能な, 特徴空間における分布が複雑に入り組み, 線形分離が不可能な場合は誤ったクラスタ結果を露呈してしまう。非線形な識別境界を実現するクラスタ手法として, ここでは非線形な分離・統合法を提案している。すなわち, クラスタリングの過程においてクラスタ分布を多次元正規分布と仮定して, その分散共分散行列の当確率線 (たとえば1シグマのコンター) の長径と短径の比を調べ, それがある値 (たとえば, 1.5) 以上であるならば, 当該クラスタを2分して, 最終クラスタ数を1つ追加してクラスタリングを継続するものとした。また, その後, 分離したクラスタを最後に統合することにした。このことにより, 既存のクラスタリング手法では分離不可能な異なるクラスタを分離することができるようになった。
この提案方法を, シミュレーション画像データおよびランドサットTM画像データを用いてk平均法とクラスタ精度を比較することによって評価したところ, 提案方法の優位性が確認できた。

抄録全体を表示