[2H11-14] Physical Protection

[2H13] Game Engine Based Data Augmentation for Malicious Behaviors Identification in Nuclear Security

*Xingyu Song¹, Zhan Li¹, Shi Chen¹, Kazuyuki Demachi¹ (1. UTokyo)

Keywords:nuclear security, malicious behaviors identification, deep learning, game engine

Nuclear security is potentially seriously compromised by external and internal threats to nuclear facilities. However, commonly employed detection systems rely on manual monitoring and extensive sensor deployment, which is time-consuming and costly. To this end, we introduce deep learning for vision-based automatic malicious behaviors identification. Given that existing benchmark datasets lack actions and characters related to nuclear security, we propose to carry out data augmentation with game engine. We customize the buildings and interiors in the adopted game engine with a 3D map editor to simulate the environment of nuclear facilities. Characters with various clothes and appearances can perform multiple actions in the simulated nuclear facility in the game engine, recording by the in-game camera in arbitrary distance and angle.