Bio-Inspired Motion Perception for Moving Object Segmentation

Motion target segmentation is a fundamental yet challenging task in computer vision. Inspired by biological vision principles, this study introduces a novel visual motion perception model for moving object segmentation, emulating the motion processing observed in the primary visual cortex (V1) and middle temporal area (MT). The model's performance was rigorously assessed using the CDnet2014 and MOT2015 datasets through an advanced background subtraction methodology. Comparative experiments with the Gaussian Mixture Model (GMM) and ViBe algorithm show that our approach achieves significantly higher precision and recall in foreground extraction. This highlights its superior ability to segment motion targets from complex backgrounds, advancing the state of the art in computer vision-based motion analysis.