A novel single-ship image deraining method for complex natural environments

With the continuous development of our nation’s water transportation industry, ensuring the safety of vessel navigation and enhancing the management of water traffic have become critical components in achieving the strategy of becoming a maritime power. Vessel target detection, as one of the core technologies in intelligent shipping, is widely applied across various fields, including maritime search and rescue, and situational assessment. However, under rainy conditions, the quality of images captured by cameras deteriorates significantly, leading to considerable loss of details. This poses a challenge to vessel target detection and greatly affects navigational safety. To improve the quality of vessel images in rainy scenarios, this paper builds upon the DDN model by incorporating a rain residual network and a detail restoration network. This approach not only enhances the model’s ability to remove rain but also maximizes the recovery of detail information in the images. Furthermore, an ECA attention mechanism is introduced to improve the model's feature extraction capabilities across different channels, further enhancing its rain removal performance. Finally, the effectiveness of this proposed method is validated through tests on both a simulated rainy dataset and real-world datasets. In the application of ship target detection, the images processed by this algorithm can accurately present the characteristics of ships, improve the recognition accuracy, facilitate the efficient execution of tasks such as maritime search and rescue, and provide reliable technical support for intelligent shipping.