Image-free dual-target single-pixel tracking

Image-free single-pixel tracking significantly reduces data throughput and computational load, but existing methods remain largely limited to single-target scenarios. To address this issue, we propose a dual-target image-free tracking method that requires only five modulation patterns per motion frame. The approach introduces reference-point-based geometric constraints in the measurement domain, enabling the precise retrieval of both target centroids directly from single-pixel measurements. The average localization errors in experiments with binary scenes remain within 3 pixels. It also performs reliably in grayscale complex scenes where the recovery error can be reduced by introducing an over-determined system and recovers trajectories even when two targets partially overlap. Furthermore, it supports potential extension to three-target scenarios and can even infer the number of moving targets. Therefore, this method opens up a new path for image-free multi-target tracking.