Vers l'assistance à l'exploration pertinente et réaliste d'environnements 3D très denses

A novel approach to navigating dense 3D environments, such as the intricate point cloud of Amiens Cathedral, leverages visual saliency and photometric entropy to guide virtual camera movements. The study introduces a Gaussian mixture-based model for visual saliency, which enhances the relevance and fluidity of camera control during both local framing and global exploration. This method ensures realistic movement while avoiding obstacles, tested in simulations and real-world scenarios with robotic systems. Additionally, an assisted exploration technique allows users to define zones of interest, combining hierarchical control laws to optimize navigation paths. These innovations address the challenges posed by highly detailed 3D environments, offering tools for more intuitive and effective virtual exploration.

Habibi, Z. Vers l’assistance à l’exploration pertinente et réaliste d’environnements 3D très denses. https://doi.org/10.70675/a050b969zece8z42f9z88abz47aaa4a4b866