ClearSight: Deep Learning-Based Image Dehazing for Enhanced UAV Road Patrol

Adverse weather conditions such as haze can deteriorate the quality of images captured during UAV patrols, thus affecting the effectiveness of object detection algorithms. To address this issue, this paper proposes an end-to-end image dehazing algorithm based on deep learning. Leveraging dilated convolutions and Transformers, we construct an image restoration module to adaptively learn and eliminate degradation information from the images. Furthermore, by adaptively weighting and fusing features from different layers of the network, we enhance the expression capability of image content features, effectively removing haze while preserving image content information. To validate the effectiveness of our approach, we also construct a dataset of foggy and clear images of UAV scenes using generative adversarial learning. Experimental results demonstrate that our proposed method can effectively remove haze from UAV images and significantly improve the accuracy of object detection in haze conditions, thereby enhancing the reliability of UAV patrols.