Yafang Hou’s research while affiliated with Northeast Normal University and other places

At present, most of the methods for urban functional areas identification are based on the road network and the types of land utilization, and cannot reflect the dynamic changes of the coverage areas and the functionalities of functional areas, accompanying with the changes of human activities. In this paper, we propose a method to identify the urban regions and analyze their spatial-temporal features based on trajectory data mining and POIs (point of interest) semantic analysis. Through taking the correlation between vehicle running conditions and func-tions of regions into account, the characteristic points in trajectory data are clustered adaptively based on their densities. The functional areas are divided reasonably through building Voronoi diagrams based on the cluster centers. In order to effectively evaluate the compositional functions of the regions, the topic words are mined and the corresponding probabilities are calculated based on the POIs’ categories in each region, using LDA (latent Dirichlet allocation) topic model. Furthermore, we propose a quantifiable method of computing the function strength based on the results of LDA. Moreover, based on the time-variant characteristics of trajectory data, an interactive visual analysis system called UFAVIS (urban functional areas visualization) is constructed to explore the impact of human activities on the spatial-temporal patterns of the functional areas. Using the method of functional area recognition with spatial-temporal feature analysis, experimental verifications and multiple case studies of the real data in Beijing are performed. The results demonstrate that UFAVIS can effectively identify the compositional functions of urban areas and find their spatial-temporal patterns changing with the variations of human activities, which provides guidance for urban planning and policy decision. © 2018, Beijing China Science Journal Publishing Co. Ltd. All right reserved.

Uncertainty is inevitable in scientific simulations. As the increase in computing power, ensemble data have been generated for multiple variables. Uncertainty has become a great challenge to the analysis of variable associations for multivariate ensemble data, as the variable associations are very complex and diverse among different ensemble members. In this paper, we propose a novel visualization method to present the uncertain associations between a reference variable and the associated variable for multivariate ensemble data. Considering the huge scale of original ensemble data, Gaussian mixture model (GMM) is exploited to quantify the uncertainty and represent the original data compactly. To reveal the spatial uncertainty of the reference variable, a GMM-based method for extracting uncertainty isosurface is proposed and shows the accuracy advantage over Gaussian-based method. Meanwhile, a data reduction method is proposed to enhance the performance of extracting uncertainty isosurface. By mapping the values of the associated variable onto the uncertainty isosurface of the reference variable, a syncretic rendering method is proposed to show the variable associations intuitively. Besides, the screen space accumulating strategy is introduced to present the uncertainties of the associations. Furthermore, we provide a switchable view for users to obtain the credibility of variable associations. The credible associations can assist users to make reliable decisions. For the regions with not credible associations, the detailed information of the associations in every ensemble member can be explored through animation for further analysis. The effectiveness of our method is demonstrated by synthetic, climate and combustion data sets.

Volume rendering is an important technique of scientific visualization that can help people analyze and understand multivariable volume data effectively. Since the previous visualization methods of multivariable volume data are not intuitive and difficult to operate, we propose a novel framework of visualizing multivariable volume data, which combines subspace clustering with Radial Coordinate Visualization (RadViz) from the global pattern analysis to the local feature exploration. Since multivariable data generally have a large data size, the feature sampling is performed to extract some representative points. In order to explore the features interactively, the sample points extracted from high-dimensional space are projected into a low-dimensional space. Through selecting different sample points interactively, users can switch and explore different subspaces in real-time. For the further analysis of the local details in the selected subspace, we utilize the RadViz technique to present the data patterns in the subspace. Thus, the relationships of the data among different dimensions can be recognized intuitively. The result of the experiment shows that our method can help users explore the complex features in volume data deeply and express the data patterns among different dimensions exactly. The constructed system based on subspace analysis and multidimensional projection visualization (SAMP-Viz) can improve the efficiency of analyzing multivariable volume data and guarantee the real-time volume rendering.

Generating large scale terrains that conform to the morphology of real scenes is a great challenge for terrain modelling, as simulating complex geometric details is time-consuming and the realistic geographical features are hard to be controlled. In this paper, we propose an efficient modeling method for large scale terrain visualization based on hydrology. To simulate real geographic features, we introduce the hydrology based Tokunaga river network to guide the terrain generation, and propose a production rule set of river network using procedural modeling. The distribution and structure of river network can be adjusted by user interactions. Ridges are extracted based on river network to provide more skeleton features, and the enrichment method of skeleton features is presented to maintain the morphology of valleys and ridges. Based on the enriched features, diffusion equation is exploited to compute the full elevation field, which can achieve the nature transitions of the regions between skeleton features. Large scale terrain with real morphological features can be generated on-line through the parallel implementation of diffusion equation. According to user requirements, the augmented virtual terrain can be obtained by blending the selected real terrain with the synthesis terrain seamlessly. Experiments are conducted on Digital Elevation Model (DEM), and the results show that the proposed methods can generate large scale terrains that conform to morphology of real terrain and can well simulate various natural scenes.

Correlation analysis is one of the most important tasks in the field of visualization research and data mining. This paper proposes a novel dissimilarity-preserving cluster algorithm that characterizes not only the time-varying patterns but also the spatial positions to summary the correlation connection in multi-variable and time-varying data sets. A temporal multi-variable structure is defined to express temporal information of a voxel in multi-dimensional space. Furthermore, a method based on structural similarity index measurement is proposed to compute the difference of time-varying pattern. In order to further explore some abnormal phenomena, spatial similarity is embedded as spatial distance metric by building the kernel density estimate for the neighborhood of each voxel. To verify the effectiveness of the method, the voxels are classified based on the time-varying similarity and spatial distance. Moreover, the combinations of two metrics are rebalanced to be suitable for the different datasets. The approach proposed in this paper is used on both synthetic and real-world data sets to demonstrate its usefulness and effectiveness.