Machine Vision Technology Empowers AI-Powered Ship Target Recognition Algorithms for Maritime Navigation

View at publisher Summarise with VIRAT
Authors: Xiandong Zhao, Yonggang Zong
Cited by: 0
References: 3

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

To address the challenges of fog, rain, sudden changes in nighttime illumination, and low recognition accuracy for small ship targets in maritime scenarios, a machine vision ship target recognition algorithm based on the Sea State Adaptive Dynamic Attention Network (HADAN) is proposed. This algorithm utilizes a three-layer architecture: multimodal preprocessing, sea state-adaptive feature extraction, and multi-scale fusion recognition. First, the algorithm optimizes visible and infrared image quality through improved CLAHE and wavelet transforms. A dynamic attention module is then constructed, combining real-time visibility and wave height parameters. Feature weights are dynamically adjusted to suppress background interference. Finally, multi-scale feature fusion is used to enhance the representation of small ship targets. Experiments conducted on a dataset of 12,000 multimodal images collected in the Bohai and Yellow Seas (covering four scene types: clear, foggy, rainy, and night), comparing ResNet50, ViT, and FusionNet algorithms, revealed that HADAN achieved a test set accuracy of 96.8%, a 7.5 percentage point improvement over ResNet-50. In foggy conditions (visibility ≤ 500m), the accuracy reached 93.5%, with a small target miss rate of only 4.8%, outperforming FusionNet by 4.2 and 4.7 percentage points, respectively. The inference speed reached 28.6 FPS, meeting the real-time requirements of maritime navigation. The experiments demonstrated that HADAN effectively addresses the challenges of ship recognition in complex sea conditions and provides reliable support for target perception in autonomous maritime AI systems.

Fields of Study & Subjects

Advanced Neural Network Applications Advanced Image Fusion Techniques Maritime Navigation and Safety

Publication details

Journal / source
IEEE International Conference on Power, Electronics and Green Energy (ICPEGE)
Pages
1389–1394
Publication date
28 January 2026
Language
EN