Multi-wavelength single-pixel non-line-of-sight imaging with a compressive sensing measurement matrix

Non-line-of-sight (NLOS) imaging aims to reconstruct objects obscured by direct line of sight. Traditional Single-pixel Imaging (SPI) performs correlation operations on signals through the illumination pattern and intensity of a single-pixel detector. However, the reconstructed result mainly provides spatial information of objects, which limits its practical applications, including autonomous driving and smart cities for defense. In this work, leveraging active correlations-based imaging techniques, a multi-wavelength single-pixel non-line-of-sight (NLOS) reconstruction framework is proposed. By introducing compressive sensing, a Total Variation minimization (TV) RGB color space algorithm is designed for more object information reconstructions via under-sampling. The proposed approach is capable of reconstructing both the space and color information of hidden objects with fine detail under the intermediate reflector and filter settings. The experimental results demonstrate that the proposed scheme achieves a compression rate of 29% and outperforms conventional single-pixel imaging in terms of object information at low sampling rates, having potential practical applications.