Design and implementation of an AR sensory fruit cutting system based on monocular video
Wei Lai, Tingsong Lu, Haidi Bi · 2024
With the rapid development of motion sensing technology, the interactive experience of motion-controlled games based on pose estimation is becoming increasingly popular. The key to enhancing these experiences lies in accurately estimating human body poses. This paper aims to explore the research on interactive motion-controlled games based on human body pose estimation. We propose a novel approach that enables real-time and smooth 3D pose estimation solely through a single-camera capture. The proposed algorithm has been applied to augmented reality games, thereby revolutionizing the traditional touch-based game interaction mode. This new form of interaction brings players an immersive gaming experience, which represents a significant innovation in the gaming industry and opens up new possibilities for future augmented reality games. We introduce a data smoothing algorithm to improve the output results of 3D pose estimation. By applying data smoothing techniques, we successfully reduce the jitter and abrupt fluctuations in pose estimation results, thereby enhancing the accuracy and stability of pose estimation. This technology provides a solid foundation for the interactive experience in games, allowing users to effortlessly control their actions, such as slicing fruits. Furthermore, it enhances the gameplay and enjoyment, while serving as a platform for studying the impact of different interaction modes and user characteristics on the overall gaming experience.