Non-Cooperative Ship Target Fusion-Based Recognition with Deep Learning

Non-cooperative ship targets plays an important role in modern naval warfare recognition and reconnaissance. In the complicated environment, information obtained from a single source often has a high degree of uncertainty. By combining the information obtained from different sources, a more accurate recognition of the target can be expected. In this paper, we propose a fusion-based recognition method with deep learning. We build multiple classifiers based on handcraft features and deep features respectively. A meta-learning method is used to implement multiple classifiers fusion. Our experimental results show that the proposed method provides better recognition accuracies for non-cooperative ship targets in complex environments compared with single classifiers and traditional voting methods when performing multiple-class classification.