Single-Pixel Detection Based Tracking of Multiple Moving Objects Using Window Geometric Moment

Fast localization and tracking of multi-targets is of great significance in many fields such as guidance, area surveillance, and unmanned aerial vehicle detection, etc. Benefiting from the unique mechanism of single-pixel imaging, a single-pixel detection based tracking method with windowed geometrical moment detection is proposed to achieve fast multiple objects tracking. Fourier slice pattern is utilized to obtain a one-dimensional projection of the scene and design the window function, which is combined with zero-order geometric moment patterns to obtain the region of the multi-target at first. In the subsequent frame tracking, the window function of each target is combined with the three geometric moments patterns to obtain the center of mass position, and the window function is updated to realize the matching correlation, and the sampling rate can be as low as 0.0034%, which greatly reduces the sampling time to achieve fast tracking. With a modulated frame rate of 20kHz, tracking can be done at frame rates up to 2200 Hz. While the window overlaps, the window transformation, and multi-directional Fourier slice segmentation method are designed to solve the tracking error. A background subtraction method is applied in complex backgrounds. The simulations of a scene with a spatial resolution of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$512 \times 512$</tex> pixels are performed, and the results show that the proposed method tracks the localization of three targets with an error of less than one pixel in both horizontal and vertical directions.