AGCG: Air–Ground Collaboration Geolocation Based on Visual Servo With Uncalibrated Cameras

Long-range target geolocation typically requires precise camera parameters. This article presents a cooperative air–ground visual servo localization system, which is suitable for uncalibrated cameras in outdoor environments. The system controls the drone to align with targets in the image, thereby determining the target position. In situations, where only rough camera parameters are known, we propose a visual servo method based on a binocular localization model. When the parameters of the localization system are completely unknown, we introduce a parameter estimation method to obtain reliable parameters for drone control. Experimental results demonstrate that visual servo control exhibits high robustness to parameter variations, and the parameter estimation method can obtain the reliable parameters for boosting the system general applicability, even under arbitrarily given initial parameters. Real-world experiments show an average localization accuracy of 4.4 m at distances greater than 300 m.