关键词: 舰船识别, 目标检测, 多尺度表达, 多模型协同, 知识融合

Abstract: In the context of globalization, the importance of ship monitoring is becoming more and more prominent, moreover, with the continuous progress of remote sensing imaging technology, ship target detection has also become a key means to ensure the safety and efficiency of maritime transportation, which is crucial for maritime transportation, environmental protection and national security. However, due to the large-scale difference and complex background of ship targets, the existing approach relies on transitional manner, it can’t adapt to ship targets with variable scales. In this paper, we propose a multi-model co-training framework, in which multiple trained ship detection models are regarded as auxiliary, optimizing the main network training by knowledge migration. First, ternary relationship constraints are introduced to transfer knowledge between the auxiliary network and the main network, followed by a soft-label guidance strategy to further improve the accuracy of ship detection. The experimental results show that compared with the existing main-stream methods, the proposed method demonstrates better performance on DOTA and xView datasets, overcomes the limitation of a single model, and provides a new solution idea for target detection in remote sensing images.

Key words: ship recognition, target detection, multi-scale representation, multi-model collaboration, knowledge fusion