Vision-based feature detection and geometric modeling for precise inter-UAV distance measurement

Abstract The safe operation of unmanned aerial vehicle (UAV) swarms relies heavily on accurate air-to-air ranging to maintain appropriate inter UAV distances. In complex environments where communication constraints limit traditional ranging methods, vision-only techniques have become an important research direction. This paper presents a novel vision-based ranging approach designed for multi-rotor UAV swarms under a downward-looking field of view. The method exploits the geometric characteristics and auxiliary stripe features of multi-rotor UAVs, and integrates elliptical fitting with a polar angle aggregation model to establish a robust mapping between two-dimensional image features and three-dimensional spatial coordinates. The Perspective-n-Point algorithm is then employed to compute relative distances with high accuracy. Flight experiments with a 1+1 UAV swarm validate the effectiveness of the proposed approach at altitudes up to 120 m. Compared with ground truth measurements, the maximum ranging error is 2.24% in a homogeneous background and 6.78% in a heterogeneous background. In addition, the method demonstrates improved robustness over state-of-the-art multi-sensor fusion techniques when applied to vision-only scenarios. These results confirm that the proposed approach provides a reliable and efficient solution for long-distance ranging in UAV swarms, enhancing the autonomy and safety of swarm systems. The applicability of the method is currently limited to configurations without overlapping rotor arms, but it offers a promising foundation for vision-based ranging in future swarm operations.