Bio-Inspired Small Target Motion Detection With Spatio-Temporal Feedback in Natural Scenes

Area of Science:

• Computer Vision
• Bio-inspired Computing
• Neuroscience

Background:

• Detecting small, distant moving objects is difficult due to limited visual features.
• Flying insects possess advanced visual systems for tracking small targets, offering a biological template.
• Existing small target motion detector (STMD) methods struggle in complex natural environments.

Purpose of the Study:

• To develop a bio-inspired visual system for robust small target motion detection.
• To enhance the discrimination of small moving targets from background clutter.
• To address performance degradation of current STMD methods in natural scenes.

Main Methods:

• Proposed a novel bio-inspired visual system with a spatio-temporal feedback mechanism.
• Designed two complementary subnetworks: one for background movement extraction, another for small target motion.
• Integrated subnetworks via a feedback loop for recurrent filtering of false positives.

Main Results:

• The Spatio-Temporal Feedback STMD system effectively suppresses false positive background movements.
• The system demonstrates enhanced responses to small targets in cluttered environments.
• Experimental results show superior performance compared to existing methods.

Conclusions:

• The proposed bio-inspired system significantly improves small target motion detection.
• Spatio-temporal feedback is crucial for filtering background noise in complex scenes.
• This approach offers a more competitive solution for real-world applications.