Looming detection in complex dynamic visual scenes by interneuronal coordination of motion and feature pathways

Detecting looming signals for collision avoidance faces challenges in real-world scenarios due to interference from moving backgrounds. Astonishingly, animals, like insects with limited neural systems, adeptly respond to looming stimuli while moving at high speeds. Existing insect-inspired looming detection models integrate either motion-pathway or feature-pathway signals, remaining susceptible to dynamic visual scene interference. We propose that coordinating interneuron signals from the two pathways could elevate looming detection performance in dynamic conditions. We used artificial neural network (ANN) to build a combined-pathway model based on Drosophila anatomy. The model exhibits convergent neural dynamics with biological counterparts after training. In particular, a multiplicative interneuron operation enhances looming signal patterns. It reduces background interferences, boosting looming detection accuracy and enabling earlier warnings across various scenarios, such as 2D animated scenes, AirSim 3D environments, and real-world situations. Our work presents testable biological hypotheses and a promising bio-inspired solution for looming detection in dynamic visual environments.