ActiveNeuS: Neural Signed Distance Fields for Active Stereo

TL;DR

With this technique, textureless or equivalent surfaces by low light condition are successfully reconstructed even with a small number of captured images, and the proposed method could achieve state-of-the-art reconstruction quality under such severe condition.

Abstract

3D-shape reconstruction in extreme environments, such as low illumination or scattering condition, has been an open problem and intensively researched. Active stereo is one of potential solution for such environments for its robustness and high accuracy. However, active stereo systems usually consist of specialized system configurations with complicated algorithms, which narrow their application. In this paper, we propose Neural Signed Distance Field for active stereo systems to enable implicit correspondence search and triangulation in generalized Structured Light. With our technique, textureless or equivalent surfaces by low light condition are successfully reconstructed even with a small number of captured images. Experiments were conducted to confirm that the proposed method could achieve state-of-the-art reconstruction quality under such severe condition. We also demonstrated that the proposed method worked in an underwater scenario.

Figures (24)

• Figure 1: System configuration.
• Figure 2: Pipeline schematics of the proposed method. Parameters marked in red are optimized during training.
• Figure 3: Example images of the synthetic data with pattern projection. Left: NeRF-Synthetic (Normal and Dark). Right: BlendedMVS (Normal and Dark). Note that only laser curves are visible for human eyes for Dark illumination scenario.
• Figure 4: Comparison of the reconstructed shapes on NeRF-Synthetic dataset. 20 images are used for input.
• Figure 5: Comparison of the reconstructed shapes on BlendedMVS dataset. All images are used for input.