Non Line of Sight Imaging

Non-line-of-sight (NLOS) imaging aims to reconstruct images of objects hidden from direct view by using indirect light reflections. Current research heavily utilizes machine learning, particularly deep neural networks, often incorporating physical priors about light transport to improve reconstruction accuracy and generalization across diverse scenarios, including low signal-to-noise conditions and dynamic environments. These advancements are driven by the potential for NLOS imaging to enhance autonomous navigation, improve robotic perception in challenging environments, and enable novel applications in areas such as search and rescue. Ongoing work focuses on reducing data acquisition time and improving robustness to real-world conditions like scattering media and moving targets.