<abstract xmlns="http://www.w3.org/1999/xhtml">
<p>Technological innovation for re-identifying maritime vessels plays a crucial role in both smart shipping technologies and the pictorial observation tasks necessary for marine recon- naissance. Vessels are vulnerable to varying gradations of engaging in the marine environment, which is complicated and dynamic compared to the conditions on land. Fewer picture samples along with considerable similarity are characteristics of warships as a class of ship, making it more challenging to recover the identities of warships at sea. Consequently, a convolutional dynamic alignment network (CoDA-Net) re-identification framework is proposed in this research. To help the network understand the warships within the desired domain and increase its ability to identify warships, a variety of ships are employed as origin information. Simulating and testing the winning of war vessels at sea helps to increase the network’s ability to recognize complexity so that users can better handle the effects of challenging maritime environments. The impact of various types of ships as transfer items is also highlighted. The research results demonstrate that the enhanced algorithm increases the overall first hit rate (Rank1) by approximately 5.9%; it also increases the mean average accuracy (mAP) by approximately 10.7% and the correlation coefficient by 0.997%.</p>
</abstract>

Technological innovation for re-identifying maritime vessels plays a crucial role in both smart shipping technologies and the pictorial observation tasks necessary for marine recon- naissance. Vessels are vulnerable to varying gradations of engaging in the marine environment, which is complicated and dynamic compared to the conditions on land. Fewer picture samples along with considerable similarity are characteristics of warships as a class of ship, making it more challenging to recover the identities of warships at sea. Consequently, a convolutional dynamic alignment network (CoDA-Net) re-identification framework is proposed in this research. To help the network understand the warships within the desired domain and increase its ability to identify warships, a variety of ships are employed as origin information. Simulating and testing the winning of war vessels at sea helps to increase the network’s ability to recognize complexity so that users can better handle the effects of challenging maritime environments. The impact of various types of ships as transfer items is also highlighted. The research results demonstrate that the enhanced algorithm increases the overall first hit rate (Rank1) by approximately 5.9%; it also increases the mean average accuracy (mAP) by approximately 10.7% and the correlation coefficient by 0.997%.

Re-Identifying Naval Vessels Using Novel Convolutional Dynamic Alignment Networks Algorithm

Department of Computer Science and Engineering

Polish Maritime Research

This work is licensed under the Creative Commons Attribution 4.0 International License.

Technological innovation for re-identifying maritime vessels plays a crucial role in both smart shipping technologies and the pictorial observation tasks...

Re-Identifying Naval Vessels Using Novel Convolutional Dynamic Alignment Networks Algorithm

Published Online: Mar 29, 2024

Page range: 64 - 76

DOI: https://doi.org/10.2478/pomr-2024-0007

Keywords
Warship re-identification, Transfer learning, Domain adaptability, CoDA-Net, Con-voluted Sea environment

© 2024 Sudipta Roy et al., published by Sciendo

This work is licensed under the Creative Commons Attribution 4.0 International License.

Re-Identifying Naval Vessels Using Novel Convolutional Dynamic Alignment Networks Algorithm

Published Online: Mar 29, 2024

Page range: 64 - 76

DOI: https://doi.org/10.2478/pomr-2024-0007

KeywordsWarship re-identification, Transfer learning, Domain adaptability, CoDA-Net, Con-voluted Sea environment

© 2024 Sudipta Roy et al., published by Sciendo

This work is licensed under the Creative Commons Attribution 4.0 International License.

Keywords
Warship re-identification, Transfer learning, Domain adaptability, CoDA-Net, Con-voluted Sea environment