To address the challenges of high cost, vulnerability to interference, and system complexity associated with traditional positioning methods in UAV swarm flight within complex environments, this paper proposes a monocular vision measurement method based on companion flight imagery. By constructing a ″Rigid Baseline-to-Image Ratio Visual Ranging Model″, the wingtip line is used as a geometric benchmark to calculate relative position through optical projection transformation, while incorporating carrier attitude information for dynamic correction. Test results demonstrate that the ground simulation achieves 3D measurement errors of less than 2 cm within a 2 to 7 meter range. Flight tests within the operational envelope (X: 13 to 30 m, Y: 0 to 2.5 m, Z: 0 to 4 m) show root mean square errors better than 0.66 m, 1.16 m and 0.78 m in the three directions respectively, with a processing speed of 20 fps and stable performance across various formation configurations. This lightweight vision measurement technology effectively resolves the conflict between feature extraction quality and real-time performance for moving targets, providing a reliable technical solution for autonomous coordination of UAV swarms in complex electromagnetic environments.