To address the challenges of maritime and aerial target recognition under adverse conditions such as rain-fog interference, wave glare, and harsh weather, this paper proposes a detection framework integrating image enhancement and deep learning optimization. First, a dehazing algorithm based on guided filtering and atmospheric scattering models is constructed to resolve image degradation in low-visibility scenarios. Second, a sea-sky line detection module combining local Otsu segmentation and Hough transform is designed to achieve precise division of sea-sky regions through adaptive threshold segmentation, thereby narrowing the target search range. Finally, the ACMix hybrid convolution module is introduced into the YOLOv8 algorithm to enhance feature fusion capabilities, while a multi-scale detection head mechanism improves sensitivity to small and weak maritime targets. Experimental results demonstrate that the improved algorithm significantly enhances average precision (AP) and detection speed in complex environments such as rain-fog and salt fog crystallization. Compared to baseline models, it exhibits clear performance advantages, balancing recognition accuracy and real-time requirements, and provides a reliable technical solution for maritime security and emergency rescue.