Insects can achieve rapid and precise collision detection despite having limited neural resources. This efficiency provides a vital reference for the development of artificial collision detection systems. Existing bio-inspired models typically include LGMD-based and correlation-based methods. Methods in the former category suffer from a non-linear dependency of warning time on the object’s contrast against the background due to the strong reliance on inter-frame intensity differences. While the latter effectively describe motion perception by leveraging local motion information derived from a delay-and-correlation mechanism, they lack precise spatial boundaries, failing to isolate the actual moving target across irrelevant background dynamics. In this paper, we propose a bio-inspired visual system with a motion-contour-guided mechanism to suppress false-positive background movement while achieving contrast-independent looming warning generation. Specifically, the proposed visual system is composed of two synergistic pathways. The first pathway is designed to extract motion cues and spatial perception of motion via neuronal ensemble coding, whereas the second pathway is developed to extract the contour of the moving target by employing geometric contour evolution. By integrating this derived contour with localized motion cues, the system analyzes the dynamic evolution of the target’s boundary to identify potential collision threats. Benefiting from this fusion of structure and motion, experimental results demonstrate that the proposed visual system is more robust than conventional bio-inspired models in collision detection across distinct contrast scenarios.
Yao et al. (Sat,) studied this question.