Presentation
Modeling Human Tasks and Motion in 360-Degree Videos for Real-Time Digital Twin Applications
DescriptionThis study aims to develop a real-time framework that uses 360-degree video to estimate human posture and dynamically detect task transitions. Unlike traditional motion tracking systems that rely on wearable sensors or single-camera setups, the proposed method integrates markerless human posture estimation with task classification and digital twin visualization. Raw video frames are processed using MediaPipe, which detects and tracks 33 skeletal landmarks per frame. Kinematic features—such as velocity, acceleration, and joint displacement—are extracted, and task transitions are classified using Gaussian Mixture Models. These data are integrated into a digital twin framework using Gaussian Splatting for smooth and continuous rendering of posture changes. The system effectively captured movements and classified them into task-specific segments. Results demonstrated that posture-based task detection can improve workplace safety by identifying inefficient or hazardous postures. This scalable, non-intrusive approach provides valuable data for ergonomic assessment and safety training.
Event Type
Lecture
TimeThursday, October 16th5:10pm - 5:30pm CDT
LocationGrand A
System Development
Similar Presentations