Foreground-background separation based on homologous signal matching and superposition in motion camera scenes

To address the challenge of foreground target detection caused by dynamic background displacement in sports camera scenarios, this paper proposes a robust separation algorithm based on homologous signal matching and multi-frame superposition. By introducing the SURF (Speeded-Up Robust Features) algorithm for rapid feature matching between adjacent frames, we estimate camera motion parameters and construct a reverse motion compensation model to align dynamic backgrounds to a global static coordinate system. Combined with multi-frame weighted accumulation optimization for background reconstruction to suppress motion target ghosts, precise foreground extraction is achieved through adaptive background differencing. This research provides a lightweight solution for real-time target detection under sports camera conditions, with application potential in intelligent security and unmanned inspection fields. Future work will further explore multimodal data fusion and hardware acceleration optimization.