Efficient Camera Path Estimation for In-Situ Visualization Using Adaptive Time-step Sampling

• Published in: Transaction of the Japan Society for Simulation Technology Vol. 16; no. 2; pp. 60 - 71
• Main Authors: Bi, Chongke, Iwata, Ken, Nonaka, Jorji, Matsushima, Taisei, Sakamoto, Naohisa, Adachi, Kazuya
• Format: Journal Article
• Language: Japanese
• Published: Japan Society for Simulation Technology 2024; 日本シミュレーション学会
• Subjects: camera path estimation; CFD simulation; CFDシミュレーション; in-situ visualization; in-situ可視化; information entropy; Kullback-Leibler divergence; カメラ移動経路推定; カルバック・ライブラー情報量; 情報エントロピー
• ISSN: 1883-5031; 1883-5058
• DOI: 10.11308/tjsst.16.60

For large-scale numerical simulations in which data I/O problems emerge as real challenges, there is a growing demand for in-situ visualization, which is expected to significantly reduce data I/O costs by performing visualization simultaneously with simulation. However, in-situ visualization outputs images rendered with pre-defined visualization parameters, thus causing lack of interactivity in visual exploration during the analysis phase. To solve this problem, this work focused on the smart in-situ visualization approach, which tries to automatically determine ‘when’ (time-step) and ‘where’ (viewpoint) to visualize during the simulation run. Although this approach can evaluate and visualize an optimal viewpoint from multiple viewpoints, the cost of evaluating the optimal viewpoint proportionately increases as the number of viewpoints increases. In this paper, we propose a method to reduce the cost of estimating the optimal viewpoint by evaluating and selecting important time domains as the simulation evolves. We also propose a method to increase the efficiency of estimating the camera movement path, which connects the obtained optimal viewpoints. In the experiments, the proposed method was applied to an oral airflow simulation, and its effectiveness was evaluated.