臺灣博碩士論文加值系統

合成孔徑雷達(Synthetic Aperture Radar，SAR)是擁有優越穿透性的雷達，發射能量到達地表後反射，再由衛星接收，相較可見光，可輕易穿透雲層之遮蔽，不受天氣狀況影響，可全天候及大範圍的物件監測，且可以產生高解析度影像，目前已普遍在航空及太空載具上使用。本研究取得SOSD資料集（Ship and Oil Spill Dataset），並進行數據增強增加訓練樣本，提高偵測準確率，另外也取得SSDD（SAR Ship Detection Dataset）資料集，能夠進行進一步的驗證與比較，藉此評估模型的效能。對船隻目標偵測之深度學習方法為YOLOv3(You Only Look Once version 3)，相較於YOLOv2，使用了多尺寸特徵融合的方式，雖增加平均偵測時間，但在小尺寸之目標有更好的精度，平均精度從94%提升至97%。與傳統圖形識別演算正確率70%相比，正確率大幅增加，平均偵測速度也提升8倍以上，並提出Downsized-YOLOv3網路架構，結果顯示此演算法準確率些許下降，但縮短約一半偵測速度，成功在衛星遙測領域中發揮其優異的性能。

Synthetic aperture radar (SAR) is a radar with superior traversal. The radar emits energy, then get the reflects after reaches the surface. Compared with visible light, it can easily penetrate the clouds and is not affected by climate conditions. SAR has a wide range of object detection and monitoring, and produce high-resolution images, also has been widely used in aviation and spacecraft.
This study obtained the ship and oil spill dataset(SOSD) and SAR ship detection dataset(SSDD). We enhanced the training samples to improve the detection accuracy.
These two datasets enable further verification and comparison.
The deep learning method for ship target detection we use is YOLOv3 (You Only Look Once version 3). Compared with YOLOv2, multi-size feature fusion is used so the average detection is added. Although cost time is increased a little bit, but the target for small size has a better accuracy, which increased from 94% to 97%.

摘要 i
ABSTRACT ii
誌謝 iv
目錄 v
圖目錄 vii
表目錄 ix
第1章 第一章 序論 1
1.1 船隻監測簡介 1
1.2 衛星影像處理方法簡介 3
1.3 研究動機與目的 4
1.4 論文內容大綱 5
第2章 第二章 物件偵測文獻回顧 6
2.1 圖形識別法 6
2.1.1 應用階層式影像切割技術於SAR影像偵測之研究 6
2.2 深度學習法 7
2.2.1 R-CNN與Fast R-CNN 9
2.2.2 Faster R-CNN 10
2.2.3 YOLO 12
2.2.4 YOLOv2 18
2.2.5 RetinaNet 27
第3章 第三章 研究方法 30
3.1 船隻影像資料集 30
3.1.1 SSDD船隻影像資料集 31
3.1.2 SOSD船隻影像資料集 31
3.2 資料前處理 33
3.2.1 影像標註 33
3.2.2 數據增強 34
3.3 YOLOv3 35
3.3.1 YOLOv3的改進 39
3.3.2 損失函數 41
3.3.3 網路縮減 41
3.4 模型評估 42
第4章 第四章 實驗結果 44
4.1 比較影像解析度差異 46
4.2 不同物件偵測方法 46
4.3 網路縮減偵測結果 47
4.4 不同資料集 50
第5章 第五章 結論與未來發展 51
5.1 結論 51
5.2 未來發展 51
參考文獻 52

[1]科學線上，海洋偵測大隊，[online]http://science-on-line.blogspot.com/2013/11/blog-post_3568.html#.XRywpugzaUk。
[2]徐逸祥、朱子豪，光學式衛星影像雲層處理之研究，航測及遙測學刊，第十七卷，第2期，第115-134頁，台北，2013。
[3]Bentz, C. and F.P.d. Miranda, "Application of remote sensing data for oil spill monitoring in the Guanabara Bay, Rio de Janeiro, Brazil," International Geoscience and Remote Sensing Symposium, Vol. 1, pp. 333-335, July 2001.
[4]Ouchi, K., "Introduction of ship detection technology using SAR and trend of ship classification," International Symposium on Ensuring Stable Use of Outer Space, March 2017.
[5]Girshick, R., J. Donahue, T. Darrell, and J. Malik, "Rich feature hierarchies for accurate object detection and semantic segmentation," IEEE International Conference on Computer Vision and Pattern Recognition, pp. 580-587, 2014
[6]Girshick, R., "Fast R-CNN," IEEE International Conference on Computer Vision (ICCV), pp. 1440-1448, December 2015.
[7]Ren, S., K. He, R. Girshick, and J. Sun, "Faster R-CNN: Towards real-time object detection with region proposal networks," IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol 39, Issue 6, pp. 1137-1149, June 2017.
[8]Redmon, J., S. Divvala, R. Girshick, and A. Farhadi, "You only look once: unified, real-time object detection," IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 779-788, December 2016
[9]Redmon, J. and A. Farhadi, "YOLO9000: Better, faster, stronger," IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 6517-6525, November 2017.
[10]T. Lin, P. Goyal, R. B. Girshick, K. He, and P. Dollar, "Focal loss for dense object detection," IEEE ICCV, 2017.
[11]Joseph Redmon and Ali Farhadi, "Yolov3: An incremental improvement," CoRR, abs/1804.02767, 2018.
[12]蕭志宇，基於深度學習合成孔徑雷達影像船隻偵測，碩士論文，國立台北科技大學電機工程系，台北，2018。
[13]Li, J., C. Qu, and J. Shao, "Ship detection in SAR images based on an improved faster R-CNN," SAR in Big Data Era: Models, Methods and Applications (BIGSARDATA), pp. 1-6, November 2017.
[14]陳威霖，應用階層式影像切割技術於SAR影像船舶和油污偵測之研究，碩士論文，國立海洋大學通訊與導航工程學系，基隆，2018。
[15]Uijlings, J.R.R., K.E.A. van de Sande, T. Gevers, and A.W.M. Smeulders, "Selective search for object recognition," International Journal of Computer Vision, Vol.104, Issue 2, pp. 154-171, September 2013.
[16]Wismann, V., M. Gade, W. Alpers, and H. Huhnerfuss, "Radar signatures of mineral oil spills measured by an airborne multi-frequency multi-polarization microwave scatterometer," Proceedings of OCEANS '93, Vol.2, pp. II348-II353, October 1993.
[17]Simonyan, K. and A. Zisserman, "Very deep convolutional networks for large-scale image recognition," arXiv preprint ,arXiv:1409.1556v6, April 2015.
[18]Jiayi Guo, Bin Lei, Chibiao Ding, Yueting Zhang, "Synthetic aperture radar image synthesis by using generative adversarial nets," IEEE Geoscience and Remote Sensing Letters, Vol. 14 , pp. 1111-1115, July 2017.