政大機構典藏-National Chengchi University Institutional Repository(NCCUR):Item 140.119/119236

政大典藏 > College of Informatics > Executive Master Program of Computer Science of NCCU > Theses > Item 140.119/119236

Please use this identifier to cite or link to this item: https://nccur.lib.nccu.edu.tw/handle/140.119/119236

Title:	基於深度學習框架之衛照圖船艦識別 Detection of Civilian Boat and War Ship in Satellite Images with Deep Learning Framework
Authors:	吳信賢 Wu, Shin-Shian
Contributors:	廖文宏 Liao, Wen-Hung 吳信賢 Wu, Shin-Shian
Keywords:	深度學習物體偵測遷移學習資料增強衛照圖資分析 Deep learning Object detection Transfer learning Data augmentation Satellite image
Date:	2018
Issue Date:	2018-08-06 18:23:57 (UTC+8)
Abstract:	本論文試圖探究如何在少量資料時利用深度學習的方法以衛照影像來進行軍民艦船的判斷，並從圖中偵測出所判別的船艦的位置。本研究使用的方法除了透過深度學習以及遷移學習(Transfer Learning) 的概念來訓練模組外，鑑於部分分類資料較少也必須用資料增強（Data Augmentation）的方式生成部分資料並加入訓練集，藉以提高整體船體偵測的準確率。經過不同模型測試與參數調校後，最佳實驗結果得出軍艦AP為0.816、民船AP為0.908，整體mAP為0.862，期許這樣的成果可進一步發展為可輔助軍事判圖人員進行相關作業之系統，提升整體作業效率。未來以此為基礎希望可以發展更細部的軍事設施判斷模組，進而可以投入其餘判圖種類之應用，完善其判圖系統。 The objective of this thesis is to develop methods to detect and recognize civilian boats and war ships in satellite images based on deep learning approaches when only limited amount of data are available. The concept of transfer learning is employed to take advantage of existing models. Owing to the restricted availability of certain categorical data, this thesis also used data augmentation techniques to generate and add samples into the training sets to improve the overall accuracy of ship detection. After extensive model selection and parameter fine-tuning, the average precision (AP) of war ships and civilian boats has reached 0.816 and 0.908 respectively, and the overall mAP is 0.862. The developed framework is ready to be incorporated in a semi-automatic system to assist military personnel in facilitating the efficiency of image detection and interpretation. This thesis is expected to lay the groundwork for more precise military facility detection models, thus improving efficacy of future military facility image detection systems.
Reference:	[1] ImageNet Large Scale Visual Recognition Challenge form http://www.image-net.org/challenges/LSVRC/ [2] Krizhevsky, Alex, Ilya Sutskever, and Geoffrey E. Hinton. "Imagenet classification with deep convolutional neural networks." Advances in neural information processing systems. 2012. [3] The PASCAL Visual Object Classes form ttp://host.robots.ox.ac.uk/pascal/VOC/ [4] ImageNet data set from http://image-net.org/ [5] Cocodataset form http://cocodataset.org/#home [6] Ren, Shaoqing, et al. "Faster r-cnn: Towards real-time object detection with region proposal networks." Advances in neural information processing systems. 2015. [7] He, Kaiming, et al. "Mask r-cnn." Computer Vision (ICCV), 2017 IEEE International Conference on. IEEE, 2017. [8] Liu, Wei, et al. "Ssd: Single shot multibox detector." European conference on computer vision. Springer, Cham, 2016. [9] Redmon, Joseph, et al. "You only look once: Unified, real-time object detection." Proceedings of the IEEE conference on computer vision and pattern recognition. 2016. [10] Lin, Tsung-Yi, et al. "Focal loss for dense object detection." arXiv preprint arXiv:1708.02002 (2017). [11] Girshick, Ross. "Fast r-cnn." arXiv preprint arXiv:1504.08083 (2015). [12] Redmon, Joseph, and Ali Farhadi. "YOLO9000: better, faster, stronger." arXiv preprint (2017). [13] YOLO v2 from https://www.youtube.com/watch?time_continue=3&v=VOC3huqHrss [14] Ma, Zhong, et al. "Satellite imagery classification based on deep convolution network." Int. J. Comput. Autom. Control Inf. Eng 10 (2016): 1055-1059. [15] Albert, Adrian, Jasleen Kaur, and Marta C. Gonzalez. "Using convolutional networks and satellite imagery to identify patterns in urban environments at a large scale." Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. ACM, 2017. [16] Van Etten, Adam. "You Only Look Twice: Rapid Multi-Scale Object Detection In Satellite Imagery." arXiv preprint arXiv:1805.09512 (2018). [17] You Only Look Twice — Multi-Scale Object Detection in Satellite Imagery With Convolutional Neural Networks from https://medium.com/the-downlinq/you-only-look-twice-multi-scale-object-detection-in-satellite-imagery-with-convolutional-neural-38dad1cf7571 [18] Google Earth from https://earth.google.com/web/ [19] arcgis-earth from http://www.esri.com/software/arcgis-earth [20] ESRI from https://www.esri.com/en-us/home [21] labelImg from https://github.com/tzutalin/labelImg [22] Tensorflow from https://www.tensorflow.org/ [23] Keras from https://github.com/keras-team/keras [24] Pan, Sinno Jialin, and Qiang Yang. "A survey on transfer learning." IEEE Transactions on knowledge and data engineering 22.10 (2010): 1345-1359. [25] Huang, Jonathan, et al. "Speed/accuracy trade-offs for modern convolutional object detectors." IEEE CVPR. 2017. [26] Chen, Liang-Chieh, et al. "Rethinking atrous convolution for semantic image segmentation." arXiv preprint arXiv:1706.05587 (2017). [27] Goodfellow, Ian, et al. "Generative adversarial nets." Advances in neural information processing systems. 2014.
Description:	碩士國立政治大學資訊科學系碩士在職專班 104971022
Source URI:	http://thesis.lib.nccu.edu.tw/record/#G0104971022
Data Type:	thesis
DOI:	10.6814/THE.NCCU.EMCS.004.2018.B02
Appears in Collections:	[Executive Master Program of Computer Science of NCCU] Theses

Files in This Item:

File	Size	Format
102201.pdf	3210Kb	Adobe PDF2	2420	View/Open

社群 sharing

Loading...