It has become easier to acquire high-resolution images with a spatial resolution of several centimeters using a camera mounted on an unmanned aerial vehicle (UAV). Since increasing spatial resolution causes inhomogeneity in the observation target, land cover classification with spectral information affects accuracy. It may be possible to improve the accuracy of land cover classification by taking texture information into account rather than using only single-pixel spectral data.
This paper proposes a novel method for improving the accuracy of land cover classification by using texture information derived from the gray level co-occurrence matrix (GLCM). Our proposed method directly vectorizes GLCM elements rather than calculating ordinal texture features. Therefore, we do not have to choose which texture feature is suitable for land cover classification. We used a support vector machine (SVM) for the classifier and confirmed the characteristics of accuracies by changing the number of gray levels (2 to 16) and the size of the calculating window (3 to 21).
In a comparison of our proposed method with simple spectral information and with spectral information with GLCM texture features, our method was found to outperform land cover classification accuracy in most cases.
Increasing human and economic losses due to urban floods demand rapid flood monitoring using synthetic aperture radar (SAR). In a global first, this study conducted simultaneous experiments using a flood experimental field that can reproduce the conditions of submerged buildings and satellite monitoring using the L-band SAR aboard the Advanced Land Observing Satellite-2 (ALOS-2). Through these experiments, we investigated the relationships among the threshold of interferometric coherence, the accuracy of urban flood detection, the multi-look number in interferometric processing, and floodwater depth. To achieve a better understanding of our experimental results, we also performed theoretical coherence simulations. Our results revealed that the coherence and flood detection accuracy statistically depends on the multi-look number and that 3×3 looks are needed to obtain reasonable accuracy. We also found that coherence-based change detection can detect urban floods with a depth of as little as 6 cm. There was no clear correlation between coherence and water depth. We also performed urban flood detection using ALOS-2 data from observed flood events; the results proved the validity of our theory and its applicability to actual disaster activities. Our findings enable robust urban flood monitoring and contribute to disaster prevention and mitigation.
Satellite remote sensing is the most cost-effective method for observing forests in areas where it is difficult to conduct field surveys, and it is expected to be used for tree species classification. Although seasonal characteristics are useful clues for tree species classification by remote sensing technology, there are few studies on tree species discrimination by satellite using these characteristics. In this study, using Tsukuba City and Boso Peninsula as test sites, we show that two evergreen broad-leaved tree species, namely, C. sieboldii and L. edulis, can be extracted from Sentinel-2/MSI images by their flowering signals. Although these two species have similar flowering signals, they can be separated from each other due to their different flowering times. The results also suggest the possibility of using the flowering signal to estimate the damage from harmful insects and diseases, and the possibility of misclassification of the land cover map by the flowering signal.
A survey of case studies and recommendations for the utilization of satellite oceanographic data was conducted. We briefly introduce the following topics, based on our survey results: satellite data portal sites in Japan and other countries, satellite data analysis software and examples of its applications, examples of systems for the application of satellite data, and data site for ocean optics and a geographic information system (GIS) to assist satellite data analysis.