As a promising technique, the Neural Radiance Fields (NeRF) and neural rendering are now widely applied in novel view synthesis and scene reconstruction. In the vanilla NeRF and subsequent neural rendering methods, one important assumption of the scene is that there is only one type of light medium in the scene, hence the light rays in such methods would remain straight during rendering. However, for underwater scenes, the camera is usually placed in a waterproofing and transplant housing. The light ray path in such a scenario would be air-water or air-housing material-water, which would cause refraction and violate the basic assumptions in vanilla NeRF. To address the issue of novel view synthesis in scenes with refractive media, this paper proposes a refractive neural rendering method under flat refractive geometry. First, the distance from the origin of the light ray to the refraction plane and the normal vector are precalibrated, which is utilized to model the per-pixel refracted ray direction, and the intersection point of the refraction plane. With the refracted rays, a neural radiance field can be trained and can be used for novel view synthesis in refractive scenes. The method is validated on synthetic and real data, revealing accurate novel view synthesis of scenes under refractive surfaces from sparse multi-view images.
The path planning algorithm is important for the safety and stability of self-piloting. In this paper, target bidirectionalRRT*(TB-RRT*) path planning algorithm based on target gravity and improved metric function is proposed to address the problems of low search efficiency, high randomness, slow convergence, and unsmooth path of the rapidly-exploring random tree star (RRT*) algorithm. Firstly, the algorithm introduces the target gravity and dynamically adjusts the sampling step to improve the search efficiency of the algorithm and reduce the randomness of the search tree growth; through the bidirectional tree growth strategy, the convergence speed of the algorithm is improved. Secondly, the smoothness of the planned paths is improved by considering both Euclidean distance and pinch angle effects on path planning with an improved metric function. Finally, the smoothing path with optimal path cost is obtained by the path selection method. Simulation results show that the improved TB-RRT* algorithm reduces the path length by 12.8%, the running time by 56.9%, and the number of sampled nodes by 43.8%.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.