首页 > , Vol. , Issue () : -
Cross-domain transfer learning aims to utilize the public datasets, as the source data, to improve the recognition accuracy of target data，breaking through the limitation that the category space between source data and target data must be consistent. As for the few-shot remote-sensing ship recognition task, the existing cross-domain transfer learning algorithms have the disadvantages of transfer category restriction and negative transfer effect. Therefore, a cross-domain transfer learning algorithm based on the source data correlation sorting was proposed to solve the above problems. Firstly, the target data was added reversely into the source domain recognition task, and according to the variation of the source data recognition accuracy before and after the target data was added, the various source data were classified into strong/weak/negative correlation samples, and only the strong correlation samples would be selected. Then, the self-supervised joint learning strategy was adopted to introduce the auxiliary self-supervised angle prediction branch into the classification network in the target domain. The selected strong correlation source samples were added and only added into the training of the self-supervised branch, avoiding changing the main classification network structure. The proposed algorithm has two main advantages: firstly, the weak/negative correlation source samples are eliminated by correlation sorting, which can avoid the occurrence of negative transfer effect; Secondly, by introducing the self-supervised angle prediction branch, the information of the strong correlation source samples was fully utilized and the features with more generalization ability were extracted, while maintaining the structural integrity of the main classification network. Randomly selecting sixty-five categories samples as the source data from miniImageNet, through the comparative experiments on few-shot ship target in remote-sensing images, the following conclusions can be obtained: 1) Choosing Resnet18 as the classification network, the performance of the proposed algorithm is better than Fine-tune algorithm which is widely used in cross-domain transfer learning, and compared with the recognition algorithm which only uses the main classification network, the proposed algorithm improves the recognition accuracy of target data from 78.89% to 96.48%; 2) Using different networks to sort correlation for the source data, the selected strong correlation source samples are not exactly the same and their categories coincidence degree is close to 60%, but they are all helpful to the classification task of the target domain. At the end of this paper, through visualizing the extracted target features, it is verified that the target features extracted by using the proposed algorithm, are more abundant and have higher generalization ability.