Few-photon pixel-level target detection imaging based on 64 × 64 array GM-APD lidar system

Abstract

Geiger-mode Avalanche Photodiodes (GM-APDs) are renowned for their single-photon sensitivity and sub-picosecond time resolution, making them extensively applicable in the field of active imaging within single-photon lidar systems. Despite these advantages, detecting targets at the pixel level over long distances remains challenging due to significant interference from noise on highly sensitive detectors and a substantial decrease in target detection efficiency when the number of detection frames is reduced. To overcome these challenges, this study introduces a spatiotemporal fusion algorithm for small target detection based on Kernel Density Estimation (KDE), which enhances the detection performance of few-photon, long-distance, pixel-level targets. The algorithm utilizes adaptive bandwidth kernel density estimation tailored for pixel-level target detection and employs spatial area smoothing based on gradient distribution. To validate its effectiveness, the algorithm was implemented in a 64 × 64 array GM-APD photon-counting lidar system and underwent daytime field experiments for verification. The results demonstrated that the system could successfully detect and identify small targets with a cross-sectional area of 0.04 m² at a distance of approximately 430 m with 20 frames. Compared to conventional data processing methods, the proposed algorithm significantly improved the system’s target detection efficiency in complex backgrounds, exhibiting strong robustness and adaptability.