Supervisor:
Southwest Ordnance Industry Bureau
Organizer:
Chongqing Ordnance Industry Society
Chongqing University of Technology
Chongqing University of Technology
Infrared ship target recognition method based on deep semi supervised learning
| DOI: | 10.11809/bqzbgcxb2025.01.035 |
| Keywords: | infrared ship; target recognition; semi supervised learning; cost sensitive learning |
| Abstract: | In the process of infrared ship target recognition in complex scenarios, there is a problem of decreased recognition accuracy due to the imbalance of data set samples and incomplete labels. A deep semi supervised learning method for infrared ship target recognition is proposed based on deep semi supervised learning. Firstly, a deep semi supervised generative adversarial network is designed by combining semi supervised learning methods. Then, the false samples generated by the generator, a small number of labeled samples, and a large number of unlabeled samples are input together to the discriminator network for training. Then, a cost sensitive learning method is introduced to design a loss function to alleviate the imbalance caused by dominant samples in gradient propagation. Finally, comparative verification is conducted on the real world data set. The results show that the proposed method effectively improves the recognition performance of unbalanced infrared ship samples. |
| Issue: | Vol. 46 No. 1 (2025) |
| Published: | 2025-01-31 |