Abstract: In response to the demands of embedded object tracking systems in fields such as intelligent surveillance and modern national defense, an adaptive correlation filtering object tracking system was designed using the domestic Unigroup PG2L100H field-programmable gate array (FPGA). The system achieves target-scale adaptation through a multiscale search strategy and guides the model updating using the average peak-to-correlation intensity criterion. A time-division multiplexing correlation filtering hardware architecture, based on resource sharing and pipelining, is proposed to address FPGA resource constraints. The system includes a two-dimensional fast Fourier forward and inverse transform module and three bilinear interpolation image-scaling modules, resulting in a low-resource-consumption FPGA tracking system. With the input image resolution of 1280 × 720, the tracking system achieved real-time object tracking at 60 frames per second, demonstrating robust tracking performance in scenarios with scale variation and cluttered backgrounds.