Computer Vision-Based Visual Hovering Attitude Adjustment Control Method for Power Drones

To achieve real-time and accurate visual navigation and stable hovering of small power inspection drones, this paper proposes a computer vision based drone hovering attitude control method. This method combines Lie algebra rotation average and point line mixed feature technology, significantly improving the attitude estimation accuracy and system stability of unmanned aerial vehicles during autonomous landing. In response to the challenges posed by complex environments in power scenarios for drone landing, this paper uses the Lie algebra method to achieve efficient global rotation estimation, and combines point line mixed features to construct an objective function to optimize the attitude parameter calculation process. The experimental results show that the pixel coordinate deviation is maintained within±80 pixels at a height of 1.3 meters, corresponding to a ground distance error of approximately ± 150 millimeters. In multiple inspection areas, the maximum cruise angle deviation of the experimental group's drones was below the ideal threshold of 95°, while the control group showed significant exceedance. Experiments and comparative analysis show that the method proposed in this paper has significant advantages in positioning accuracy, anti-interference ability, and stability, especially for harsh environments such as dense equipment, complex backgrounds, and electromagnetic interference in power inspection. It can provide a reliable technical solution for unmanned aerial vehicles to achieve autonomous, accurate, and safe close range detection and landing operations in substations, power transmission lines, and other scenarios, and has clear engineering application value.