Research on adaptive optimization of unmanned aerial vehicle transmission line inspection imaging under strong and weak light conditions

This study focuses on the problem of image quality degradation caused by drastic changes in light intensity during unmanned aerial vehicle inspection of power transmission lines, and proposes a three-level adaptive optimization framework of light perception, parameter adjustment, and image enhancement. This framework precisely identifies the light conditions through a multi-feature fusion model (with an accuracy rate of 97.3% for light scene classification), dynamically optimizes the core parameters of the camera based on reinforcement learning, and completes image enhancement by combining the lightweight network LW-EnhanceNet. Experiments show that the proportion of overexposure in strong light scenarios has decreased from 35.2% to 7.8%, the signal-to-noise ratio in weak light scenarios has reached 37.5 dB (an increase of 54.9% compared to the original image), and the F1-score for defect detection has reached 89.7%; in real scenarios, the rate of image quality meeting standards has increased to 91.0%, and the inspection efficiency has increased by 24.8%. This method provides an effective solution to the imaging bottleneck in all-weather inspections, and has significant theoretical and practical significance for improving the automation level of inspections and the accuracy of defect detection.