Multichannel multicentroid motion-compensated single pixel imaging of a 2D arbitrarily moving rigid-body target.

Single-pixel imaging (SPI), distinguished by its cost-efficiency, exceptional spectral adaptability, and robust sub-Nyquist-Shannon sampling reconstruction capabilities, demonstrates transformative potential across imaging applications yet faces critical limitations in capturing arbitrarily moving targets. This work introduces a simple yet effective SPI architecture capable of simultaneous real-time tracking and high-fidelity imaging of objects undergoing unconstrained 2D planar motion, encompassing both periodic and non-periodic translational/rotational kinematics. Our methodology employs six strategically designed Fourier patterns with optimized spatial frequencies as localization markers, combined with multichannel centroid tracking, to achieve precise motion dynamics characterization. Furthermore, we develop a straightforward inverse motion-compensated reconstruction method that efficiently reconstructs images of objects subjected to composite motion. The proposed framework notably maintains reconstruction integrity even when individual detection channels experience temporary out of the field of view.