近年來台灣與中國關係愈來愈緊張，中國解放軍船艦常於台灣附近海域進行示威，使得台灣對於軍艦辨識需求提升。本篇論文之研究目的便是使用捲積神經網路(CNN)來建構一套能辨識周遭海域可能出現的大型軍艦之辨識系統。
本次實驗使用實船影像共1743張，分屬8國20艘航母或登陸作戰相關船艦，以進行資料庫構建。由於軍艦影像取得不易，因此使用資料增強的方式來擴充影像資料庫，資料擴建後的影像資料庫共8715張影像，擴建方式使用平移、對比增強、加入高斯雜訊、影像加霧等四種。本研究之船艦辨識系統以卷積神經網路來進行系統的建構，總共選用四種網路GoogLeNet(Inception-v1)、Inception-v3、ResNet-18、ResNet-101來進行訓練，並比對這四種CNN網路對於正常情況與雲霧繚繞的情況下的實船影像辨識率。研究結果顯示，本次研究之船艦辨識系統在天氣良好的情況下的辨識率可達96.3%。在雲霧繚繞的情況下的辨識率可達89.5%。

關鍵詞:卷積神經網路、軍艦影像辨識

In recent years, the relationship between Taiwan and China has become more and more tense, and the Chinese PLA ships often conduct military exercises in the waters around Taiwan, which has increased the demand for warship identification in Taiwan. The purpose of this paper is to construct a recognition system to identify possible warships in the surrounding area using convolutional neural networks.
A total of 1,743 images of actual ships were used to construct the database, which consists 20 aircraft carriers vessels related to amphibious warfare and from 8 countries. Since it is not easy to acquire the images of warships, the image database was constructed by using data enhancement methods. The expanded image database consists of 8715 images, which were expanded using three methods: panning, contrast enhancement, and adding Gaussian noise. In this research, the ship recognition system was constructed by four convolutional neural networks, GoogLeNet (Inception-v1), Inception-v3, ResNet-18 and ResNet-101. After training the four networks were compared for the recognition rate of live ship images under normal conditions and cloudy conditions, The results showed that the recognition rate of the ship recognition system in this study could reach 96% in good weather conditions. And 78.3% under cloudy conditions.

Keyword: convolutional neural network, warship image recognition

[中文]
1. 許志彬(2007),船艦辨識法之研究，國防大學中正理工學院電子工程研究所96學年度碩士論文。

2. 陳建村(1993),利用模糊特徵表示法執行船艦辨識，元智大學電機與資訊工程研究所82學年度碩士論文。

3. 黃鈺棋(2019),應用連續顏色長度特徵演算法於軍事辨識，國立中興大學資訊管理學系(所)107學年度碩士論文。

4. 曾偉銘(2018),植基於卷積神經網路技術之自動化船艦偵測與切割，國立台中科技大學資訊工程系(所) 106學年度碩士論文。

5. 鄭富元(2006),一個多方位船艦辨識系統雛型之研究，國立中興大學電機工程學系(所) 94學年度碩士論文。

[外文]
6. Chellapilla, Kumar; Puri,Sidd & Simard, Patrice.(2006). "High Performance Convolutional Neural Networks for Document Processing", Tenth International Workshop on Frontiers in Handwriting Recognition. La Baule, Centre de Congres Atlantia, France, October, 2006.

7. Carmona-Duarte, Cristina; Dorta-Naranjo, B.P.; Asensio, Alberto & Blanco del Campo, Alvaro. (2007). " CWLFM Radar for Ship Detection and Identification", Aerospace and Electronic Systems Magazine, IEEE. Vol.22. pp.22-26. 10.1109/MAES.2007.323295.

8. Fukushima, Kunihiko. (1980). "Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position", Biological Cybernetics, Vol. 36, pp.193–202, DOI: 10.1007/BF00344251.

9. He, Kaiming; Zhang, Xiangyu; Ren, Shaoqing ＆ Sun, Jian. Microsoft.(2016). "Deep Residual Learning for Image Recognition", 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 770-778, Las Vegas, NV, USA, October 2016, DOI: 10.1109/CVPR.2016.90.

10. He, Kaiming; Sun, Jian & Tang, Xiaoou (2009). "Single Image Haze Removal Using Dark Channel Prior", IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 33, No. 12, pp. 2341-2353, DOI: 10.1109/TPAMI.2010.168.

11. Krizhevsky, Alex; Sutskever, Ilya; Hinton & Geoffrey, E.Hinton (2012). "ImageNet Classification with Deep Convolutional Neural Networks", Neural Information Processing Systems 25 (NIPS 2012), Stateline, NV, USA, MAY 2017.

12. R. T. Tan.(2008). "Visibility in bad weather from a single image", 2008 IEEE Conference on Computer Vision and Pattern Recognition, pp. 1-8, Anchorage, AK, USA, June 2008, DOI: 10.1109/CVPR.2008.4587643

13. Szegedy, Christian; Liu, Wei; Jia, Yangqing; Sermanet, Pierre; Reed,Scott; Anguelov, Dragomir; Erhan, Dumitru; Vanhoucke, Vincent & Rabinovich,Andrew. Google.(2015). "Going deeper with convolutions", IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp.1-9, Boston, MA, USA, June 2015, DOI:10.1109/CVPR.2015.7298594

14. Szegedy, Christian; Vanhoucke, Vincent; Ioffe, Sergey; Shlens, Jonathon & Wojna, Zbigniew. (2015). "Rethinking the Inception Architecture for Computer Vision", 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 2818-2826, Las Vegas, NV, USA, June 2016, DOI: 10.1109/CVPR.2016.308.

15. Simonyan, Karen; Zisserman, Andrew. (2014). "Very Deep Convolutional Networks for Large-Scale Image Recognition", International Conference on Learning Representations (ICLR), San Diego, CA, USA, May, 2015.

16. Shi, Qiaoqiao; Li, Wei; Zhang, Fan; Hu, Wei; Sun, Xu & Gao, Lianru.(2018). "Deep CNN With Multi-Scale Rotation Invariance Features for Ship Classification ", IEEE Access. pp. 1-1. 10.1109/ACCESS.2018.2853620, November 2018.

17. Wang, Yuanyuan; Wang, Chao; Zhang, Hong; Dong, Yingbo & Wei, Sisi (2019). "Automatic Ship Detection Based on RetinaNet Using Multi-Resolution Gaofen-3 Imagery" Remote Sensing, Vol. 11, No. 5, pp. 531-544, DOI: 10.3390/rs11050531

18. Yann, Lecun; Bottou, Leon; Bengio, Y. & Haffner, Patrick. (1998). "Gradient-based learning applied to document recognition", Proceedings of the IEEE, Vol. 86, No. 11, pp. 2278-2324, DOI: 10.1109/5.726791.

[網路]
19. 維基百科，過度擬合overfitting示意圖(維基百科過度擬合條目) https://zh.wikipedia.org/wiki/%E9%81%8E%E9%81%A9 ，下載日期：2021年1月19日。

20. 維基百科，高斯分布（常態分布）示意圖（維基百科常態分布條目）
https://zh.wikipedia.org/wiki/%E6%AD%A3%E6%80%81%E5%88%86%E5%B8%83，下載日期：2021年1月19日。

21. MATLAB深度學習教程網站，各類CNN網路效能比較圖 https://ww2.mathworks.cn/help/deeplearning/ug/pretrained-convolutional-neural-networks.html，下載日期：2021年1月19日。