臺灣博碩士論文加值系統

近年來，由於國際情勢關係越來越緊張，各國的國防意識抬頭，對各類別軍艦的辨識成為國防重要工作。本研究之目的，在發展出一軍艦影像辨識系統。由於軍艦影片或影像因涉及國家國防軍事機密不易取得，本研究特以3D軍艦模型來模擬軍艦於海上環境影像，虛擬出一個使用無人機進行海上巡邏拍攝的環境，希望能協助我方情報單位偵查，來防止海上的敵軍船艦入侵。
本研究使用51艘分屬六大類別軍艦的3D軍艦模型來產生不同視角、高度、遠近之軍艦影像作為資料集，並以此來進行軍艦影像識別的訓練及測試，而為了找出軍艦特徵，我們使用semi-run-length方法來強化軍艦顏色、紋路、形狀特性等特徵，並以這些特徵辨別軍艦之種類，在特徵權重參數部分使用基因演算法(genetic algorithm, GA)，以達到最佳化，實驗結果顯示，在此3D軍艦模型的訓練辨識準確度率可達到97.6%，而測試辨識準確率可達到92.2%，另外本文也有對少量軍艦在海上被拍攝的真實影像做辨識，測試準確率為73.1%。

According to the relations between countries are strained and the awareness of national defense is gradually raised in recent years, the purpose of this research is to develop a warship image recognition system that will virtualize an environment with drones for maritime patrols. However, since warship films or images are difficult to obtain due to national defense military secrets, this study will use the 3D warship model to simulate the image of the warship in the sea environment, hoping to assist our intelligence units to detect and prevent the invasion of enemy ships at sea.
This research uses the 3D warship model, including 51 models from six types of ships, to generate images of different perspectives, heights, and distances as a data set so as to apply to identifying the image of the warship. In order to find the features of warship, we use the semi-run-length method to enhance the warship's color, texture, shape features and other features that can identify the type of warship. In the feature weight parameter part, a genetic algorithm is used to approach the optimization. The experimental results show that the training accuracy rate with 3D warship model can reach 97.6%, and the test accuracy can reach 92.2%. In addition, the paper also identifies the real images captured by a small number of warships at sea and the test accuracy rate is 73.1%.

目次
摘要 i
Abstract ii
目次 iii
表目次 v
圖目次 vi
第1章 緒論 1
1.1 研究背景與動機 1
1.2 研究目的 3
1.3 論文架構 3
第2章 文獻探討 4
2.1 海上監控 4
2.2 物件偵測演算法相關理論探討 4
2.2.1 深度學習相關框架探討 4
2.3 物件辨識演算法相關理論探討 5
2.4 船隻辨識與無人機相關領域研究 6
第3章 研究方法 7
3.1 軍艦影像資料庫建置. 7
3.2 研究流程 8
3.2.1 軍艦影像辨識系統 9
3.2.2 訓練集與測試集 10
3.3 影像對比增強 12
3.4 特徵提取 14
3.4.1 調色板顏色－降階 14
3.4.2 連續顏色特徵semi-run-length 15
3.4.3 旋轉不變性 18
3.4.4 軍艦辨識系統－特徵比對 19
3.5 變異數分析F統計量－特徵篩選 21
3.6 基因演算法－決定參數 24
第4章 實驗結果 28
4.1 實驗影像資料集 28
4.2 3D軍艦模型影像分類 30
4.2.1 模型影像訓練測試環境 30
4.2.2 模型影像分類結果 31
4.2.3 影像特徵提取與參數訓練 39
4.3 網路軍艦真實影像分類 41
4.3.1 網路影像訓練測試環境 41
4.2.2 網路影像分類結果 41
第5章 結論 43
參考文獻 44
附錄 48

[1]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。國立臺中科技大學資訊工程系碩士班碩士論文，台中市。 取自https://hdl.handle.net/11296/m88c2e
[2]Dalal, N., and B. Triggs. “Histograms of Oriented Gradients for Human Detection.” 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2005, doi:10.1109/cvpr.2005.177.
[3]Yeo, B.c., et al. “Scalable-Width Temporal Edge Detection for Recursive Background Recovery in Adaptive Background Modeling.” Applied Soft Computing, vol. 13, no. 4, 2013, pp. 1583–1591., doi:10.1016/j.asoc.2013.01.012.
[4]Rakibe, R. S., and B. D. Patil. “Background Subtraction Algorithm Based Human Motion Detection.” International Journal of Scientific and Research Publications, vol. 3, no. 5, May 2013, pp. 2250–3153.
[5]Mashak, Saeed Vahabi, et al. “Background Subtraction for Object Detection under Varying Environments.” 2010 International Conference of Soft Computing and Pattern Recognition, 2010, doi:10.1109/socpar.2010.5685960.
[6]Zivkovic, Z. “Improved Adaptive Gaussian Mixture Model for Background Subtraction.” 2004 17th International Conference on Pattern Recognition, 2004, doi:10.1109/icpr.2004.1333992.
[7]Aslani, S., and H. Mahdavi-Nasab. “Optical Flow Based Moving Object Detection and Tracking for Traffic Surveillanc.” International Journal of Electrical and Computer Engineering, vol. 7, no. 9, 2013.
[8]Russakovsky, Olga, et al. “ImageNet Large Scale Visual Recognition Challenge.” International Journal of Computer Vision, vol. 115, no. 3, 2015, pp. 211–252., doi:10.1007/s11263-015-0816-y.
[9]Krizhevsky, Alex, et al. “ImageNet Classification with Deep Convolutional Neural Networks.” Communications of the ACM, vol. 60, no. 6, 2017, pp. 84–90., doi:10.1145/3065386.
[10]Simonyan, K.,and A. Zisserman. “Very Deep Convolutional Networks for Large-Scale Image Recognition.”2015 IEEE Conference on Computer Vision and Pattern Recognition, 2015.
[11]Szegedy, Christian, et al. “Going Deeper with Convolutions.” 2015 IEEE Conference on Computer Vision and Pattern Recognition, 2015, doi:10.1109/cvpr.2015.7298594.
[12]He, Kaiming, et al. “Deep Residual Learning for Image Recognition.” 2016 IEEE Conference on Computer Vision and Pattern Recognition, 2016, doi:10.1109/cvpr.2016.90.
[13]Everingham, Mark, et al. “The Pascal Visual Object Classes (VOC) Challenge.” International Journal of Computer Vision, vol. 88, no. 2, 2009, pp. 303–338., doi:10.1007/s11263-009-0275-4.
[14]Redmon, Joseph, and Ali Farhadi. “YOLO9000: Better, Faster, Stronger.” 2017 IEEE Conference on Computer Vision and Pattern Recognition, 2017, doi:10.1109/cvpr.2017.690.
[15]Ren, Shaoqing, et al. “Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks.” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 39, no. 6, 2017, pp. 1137–1149., doi:10.1109/tpami.2016.2577031.
[16]Liu, Wei, et al. “SSD: Single Shot MultiBox Detector.” Computer Vision – ECCV 2016 Lecture Notes in Computer Science, 2016, pp. 21–37., doi:10.1007/978-3-319-46448-0_2.
[17]Long, Jonathan, et al. “Fully Convolutional Networks for Semantic Segmentation.” 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2015, doi:10.1109/cvpr.2015.7298965.
[18]Orstavik, M. “AirNet – A Deep Learning Approach to Extracting Building Information from Remotely Sensed Imagery.” Faculty of Engineering Science and Technology Department of Civil and Environmental Engineering, url:{ https://github.com/mathildor/TF-SegNet}, 2017.
[19]Chuang, Meng-Che, et al. “A Feature Learning and Object Recognition Framework for Underwater Fish Images.” IEEE Transactions on Image Processing, 2016, pp. 1862–1872.2., doi:10.1109/tip.2016.2535342.
[20]Zhu, Ji-Wei, et al. “An Improved Shape Contexts Based Ship Classification in SAR Images.” Remote Sensing, vol. 9, no. 2, 2017, p. 145., doi:10.3390/rs9020145.
[21]Lowe, D.g. “Object Recognition from Local Scale-Invariant Features.” 1999 7th IEEE International Conference on Computer Vision, 1999, pp. 1150-1157, doi:10.1109/iccv.1999.790410.
[22]Bay, Herbert, et al. “Speeded-Up Robust Features (SURF).” Computer Vision and Image Understanding, vol. 110, no. 3, 2008, pp. 346–359., doi:10.1016/j.cviu.2007.09.014.
[23]Rublee, Ethan, et al. “ORB: An Efficient Alternative to SIFT or SURF.” 2011 International Conference on Computer Vision, 2011, doi:10.1109/iccv.2011.6126544.
[24]Shi, Qiaoqiao, et al. “Deep CNN With Multi-Scale Rotation Invariance Features for Ship Classification.” IEEE Access, vol. 6, 2018, pp. 38656–38668., doi:10.1109/access.2018.2853620.
[25]Li, Wenkai, et al. “Integrated Localization and Recognition for Inshore Ships in Large Scene Remote Sensing Images.” IEEE Geoscience and Remote Sensing Letters, vol. 14, no. 6, 2017, pp. 936–940., doi:10.1109/lgrs.2017.2688357.
[26]Zhu, Ji-Wei, et al. “An Improved Shape Contexts Based Ship Classification in SAR Images.” Remote Sensing, vol. 9, no. 2, 2017, p. 145., doi:10.3390/rs9020145.
[27]Xu, Yongzheng, et al. “Car Detection from Low-Altitude UAV Imagery with the Faster R-CNN.” Journal of Advanced Transportation, 2017, pp. 1–10., doi:10.1155/2017/2823617.