Understanding of human dynamics has drawn attention to various areas. The wide spread of positioning technologies, such as GPS facilitates location information to be obtained with high spatio-temporal resolution as well as at low costs. By collecting individual location information in real time, monitoring of human dynamics has recently become possible and is expected to the area of dynamic traffic control. In this monitoring, detecting anomalous states of human dynamics become important. This research aims to define an anomaly detection problem of the human dynamics monitoring with time-series gridded population data and develop an anomaly detection method for this problem. According to the result of a review we have conducted, we discussed the characteristics of the anomaly detection in human dynamics monitoring and categorized our problem to a semi-supervised anomaly detection problem that detects contextual anomalies behind time-series data. We developed an anomaly detection method based on a sticky hierarchical Dirichlet process hidden Markov model, which is able to estimate the number of latent states according to the input data. Results of the experiment with synthetic data showed that our proposed method has good fundamental performance with respect to the detection rate. Through the experiments with real gridded population data, anomalies were detected when and where an actual social event had occurred.

There is a growing need for algorithms to automatically detect objects in satellite images. Object detection algorithms using deep learning have demonstrated a significant improvement in object detection performance. However, deep-learning models have difficulty in interpreting the features for inference. This difficulty is practically problematic when analyzing earth-observation images, which are often used as evidence for public decision-making. In addition, for the same reason, it is difficult to set an explicit policy or criteria to improve the models. To deal with these challenges, we introduce a feature attribution method that defines an approximate model and calculates the attribution of input features to the output of a deep-learning model. For the object detection models of satellite images with complex textures, we propose a method to visualize the basis of inference using pixel-wise feature attribution. Furthermore, we propose new methods for model evaluation, regularization, and data selection, based on feature attribution. Experimental results demonstrate the feasibility of the proposed methods for basis visualization and model evaluation. Moreover, the results illustrate that the model using the proposed regularization method can avoid over-fitting and achieve higher performance, and the proposed data selection method allows for the efficient selection of new training data.

The Japanese city presents a certain number of peculiarities in the organization of its physical space (weak zoning regulations, fast piecemeal destruction/reconstruction of buildings and blocks, high compacity, incremental reorganization). Compared to countries where urban fabrics are more perennial and easily distinguishable (old centers, modern planned projects, residential areas, etc.), in Japanese metropolitan areas we often observe higher heterogeneity and more complex spatial patterns. Even within such a model, it should be possible to recognize the internal organization of the physical city. The aim of this paper is thus to study the spatial structure of the contemporary Japanese city, generalizing on the case study of Osaka and Kobe. In order to achieve this goal, we will need to identify urban forms at different local scales (building types, urban fabrics) and to analyze them at a wider scale to delineate morphological regions and their structuring of the overall layout of the contemporary Japanese city. Several analytical protocols are used together with field observations and literature. The results, and more particularly the building and urban fabric types and their location within the Osaka-Kobe metropolitan area, are interpreted in the light of Japanese history and model of urbanization. A synoptic graphical model of an urban morphological structure based upon Osaka is produced and proposed as an interpretative pattern for the Japanese metropolitan city in general.

The present study develops a SAR-based shoreline monitoring system which consists of: (i) selection of SAR scenes suitable for shoreline detection; (ii) extraction of shoreline from the selected scenes; and (iii) estimation of the shoreline at arbitrary water level. The first model focused on ALOS-2 SAR scenes and determined fitting formulae for estimation of the backscattering coefficients at a sea and a beach surfaces. To quantify the suitability of scenes for shoreline detection, a parameter, P, was newly introduced. The value of P represents a probability at which a backscattering coefficient at a beach becomes higher than the one at a sea and it was confirmed that P reasonably indicates suitability of scenes for shoreline detection. A shoreline extraction technique was then developed based on the edge detection technique. Bilateral filter was applied to reduce the speckle noise of SAR images and wavelet transform was applied to detect the edge location. Validity of the proposed technique was confirmed through application of the present technique to Shimizu and Shonan coasts. Finally, a simple polynomial function was introduced to represent the shoreline location at arbitrary water level. Applicability of the proposed technique was also tested and validated against the detailed data obtained by image analysis.

The extensive monitoring of shorelines is becoming important for investigating the impact of coastal erosion. Satellite synthetic aperture radar (SAR) images can cover wide areas independently of weather or time. The recent development of high-resolution satellite SAR images has made observations more detailed. Shoreline extraction using high-resolution images, however, is challenging because of the influence of speckle, crest lines, patterns in sandy beaches, etc. We develop a shoreline extraction method based on the spatial pattern analysis of satellite SAR images. The proposed method consists of image decomposition, smoothing, sea and land area segmentation, and shoreline refinement. The image decomposition step, in which the image is decomposed into its texture and outline components, is based on morphological component analysis. In the image decomposition step, a learning process involving spatial patterns is introduced. The outline images are smoothed using a non-local means filter, and then the images are segmented into sea and land areas using the graph cuts’ technique. The boundary between these two areas can be regarded as the shoreline. Finally, the snakes algorithm is applied to refine the position accuracy. The proposed method is applied to the satellite SAR images of coasts in Japan. The method can successfully extract the shorelines. Through experiments, the performance of the proposed method is confirmed.