臺灣博碩士論文加值系統

空間資訊領域常運用空間資訊及觀測系統進行空間資料處理與分析，以滿足不同領域應用需求。然而面對衛星遙測大數據影像與網路上各類型的地理空間資料，人工已難以負荷分析此龐大的影像與數據資料。為有效進行遙測影像判釋與地球觀測應用，已不能用傳統方式來作業。本研究首先整合遙測影像與興趣點(Point of Interest, POI)地理空間資料庫自動建置人工智慧技術發展所必要的已標註之訓練影像。
在電腦視覺中影像分類一直是圖像識別領域的典型研究課題。近年來，使用深度學習中的卷積神經網路(Convolution Neural Network, CNN)進行影像分類已逐漸盛行，藉由神經網路自動化提升辨識能力，為圖像識別開闢了一個新的視野。
因此，本研究提出了一個基於卷積神經網路(CNN)的光學遙測影像場景分類模型，並以InceptionV3作為轉移學習(Transfer Learning)之預訓練模型，以機場為例，使用本研究完成訓練之模型對三個測試區域，共計1,296幅影像圖磚，進行有無機場之二元分類，在轉移學習中加入微調可使分類精度平均平提升77.37%、F1 score從0.02提升至0.54及F2 score從0.28提升0.74。

Geospatial information often use spatial information and observation system for spatial data processing and analysis to meet the need of applications in different fields. However, in the face of big data satellite images and various types of geospatial data on the Internet, it is hard for users to analyze such huge images and data with great load. In order to effectively conduct image interpretation and earth observation applications, it is no longer possible to work in a traditional manner. In this study, we firstly integrate the remote sensing images and POI geospatial databases to automatically generate the huge remote sensing images that are necessary for the development of artificial intelligence technology.
Image classification in computer vision has always been a fundamental research topic in the field of image recognition. In recent years, image classification using the Convolution Neural Network (CNN) in deep learning has become increasingly popular, and neural network automatization has improved recognition capabilities. This opens up a new horizon for image recognition.
Therefore, this study employed transfer learning technique, with the fine-tuning strategy on the popular InceptionV3 network. The well-trained model predicted a total of 1,296 image tiles from the three test areas, in order to assess the binary classification between airport and not airport images. With a fine-tuning strategy added, the average accuracy was improved by 77.37%. The F1 score was found from 0.17 to 0.54; the F2 score was found from 0.28 to 0.74.

目錄

誌謝 ii
摘要 iii
ABSTRACT iv
目錄 v
表目錄 vii
圖目錄 viii
1. 緒論 1
1.1 研究背景與動機 1
1.2 研究目的 2
1.3 研究架構及流程 3
2. 文獻回顧 5
3. 理論基礎 27
3.1 人工神經網路 27
3.2.1 神經元構造 28
3.2.2 神經元參數 29
3.2.3 神經網路學習過程 34
3.2 卷積神經網路 38
3.3.1 卷積 38
3.3.2 池化 39
3.3.3 全連接層 40
3.3.4 正規化 40
3.3.5 CNN運作特性 41
3.3.6 CNN優化方法 44
3.3.7 CNN經典模型 45
3.3 精度評估指標 53
4. 研究方法 58
4.1 轉移學習 58
4.1.1 轉移學習策略 58
4.1.2 轉移學習模型 60
4.2 影像圖磚 63
4.2.1 圖磚結構 63
4.2.2 Google XYZ圖磚服務標準 65
5. 實驗設計 70
6. 實驗成果與分析 81
6.1 神經網路模型訓練 81
6.2 機場目標點位資料維護 92
6.3 機場圖磚誤判原因 95
7. 結論與建議 97
7.1 結論 97
7.2 建議 98
參考文獻 99
自傳 108

[1]潘國樑，遙測概念.原理與影像判釋技術（第二版），科技圖書出版社，臺北，2009。
[2]Ng, A., “Nuts and bolts of building AI applications using Deep Learning,” Stanford University, USA, pp.1-5, 2016.
[3]Robert, M. H., Shanmugam, K., and Dinstein, I., “Textural Features for Image Classification,” IEEE Transactions on Systems, Man, and Cybernetics, Vol. SMC-3, No. 6, pp.610-621, 1973.
[4]李瑞陽、莊佳文，“運用影像分塊方法於高解析衛星影像土地利用判釋精度之研究”，航測及遙測學刋，第11卷，第4期，，第403-415頁，2006。
[5]黃國楨，“SPOT影像應用於台北縣八里鄉土地利用監測之研究”，博士論文，國立中興大學森林研究所，臺中，第1-150頁，2008。
[6]李庭誼，“結合光譜與空間特徵之高光譜影像物件分類”，碩士論文，臺灣大學工學院土木工程學系，臺北，第1-119頁，2011。
[7]Turing, A. M., “Computing Machinery and Intelligence,” Computing Machinery and Intelligence, Vo.49, pp.433-460, 1950.
[8]http://w3.uch.edu.tw/cwhuang/aihw/ai59011001.htm(2018.5.17).
[9]Samuel, A. L., “Some studies in machine learning using the game of checkers”, IBM Journal of Research and Development, Vol. 44, No.1.2, pp.206-226, 2000.
[10]林大貴，HADOOP+SPARK大數據巨量分析與機器學習整合開發實戰，博碩文化股份有限公司，新北，第10-15頁，2015。
[11]https://blog.gcp.expert/ml-1-ai-ml-deep-learning-intro/(2018.6.26).
[12]Rumelhart, D. E., Hinton G. E., and Williams, R. J., “Learning representations by back-propagating errors,” nature, Vol. 323, pp.533-536, 1986.
[13]https://medium.com/@yehjames/%E8%B3%87%E6%96%99%E5%88%86%E6%9E%90-%E6%A9%9F%E5%99%A8%E5%AD%B8%E7%BF%92-%E7%AC%AC5-1%E8%AC%9B-%E5%8D%B7%E7%A9%8D%E7%A5%9E%E7%B6%93%E7%B6%B2%E7%B5%A1%E4%BB%8B%E7%B4%B9-convolutional-neural-network-4f8249d65d4f(2018.6.26).
[14]https://itw01.com/G5B4E8Y.html(2018.6.26).
[15]https://www.ithome.com.tw/news/120019(2018.6.26).
[16]林大貴，TensorFlow+Keras深度學習人工智慧實務應用，博碩文化股份有限公司，新北，第2-8頁，2017。
[17]LeCun, Y. and Ranzato M., “Deep Learning Tutorial,” The 30th International Conference on Machine Learning, USA, pp.1-204, 2013.
[18]邵泰璋、史天元，“類神經網路於多光譜影像分類之應用”，航測及遙測學刊，第5卷，第1期，第1-15頁，2000。
[19]林文賜、周天穎、林昭遠，“「科技短文」應用監督性類神經網路於衛星影像分類技術之探討”，航測及遙測學刊，第6卷，第1期，第41-58頁，2001。
[20]Xu, W., Tian, Y., and Huang, J., “Use of artificial neural networks for estimating crop acreage from MODIS data in large-scale,” Communications, Circuits and Systems, Vol.2, pp.988-992, 2005.
[21]Neagoe, V., Strugaru, G., “A concurrent neural network model for pattern recognition in multispectral satellite imagery,” World Automation Congress, USA, pp.1-6, 2008.
[22]Sheikh, M. A. A., “A novel self-assessed approach for classification of manmade objects and natural scene images from aerial images,” Annual IEEE India Conference (INDICON), India, pp.1-7, 2011.
[23]He, K., Zhang, X., Ren, S., and Sun, J., “Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification,” IEEE International Conference on Computer Vision (ICCV), USA, pp.1026-1034, 2015.
[24]Wu, H., Zhang, H., Zhang, J., and Xu, F., “Typical Target Detection in Satellite Images Based on Convolutional Neural Networks,” IEEE International Conference on Systems, Man, and Cybernetics, China, pp.2956-2961, 2015.
[25]Lam, D., Kuzma, R., McGee, K., Dooley, S. Laielli, M., Klaric, M., Bulatov, Y., and McCord, B., “xView: Objects in Context in Overhead Imagery,” Computer Vision and Pattern Recognition, Computer Science, Cornell University Library, pp.1-16, 2018.
[26]http://image-net.org/index(2018.4.22).
[27]Den, J., Dong, W., Socher, R., Li, L. J., Li, K., and Li, F. F., “ImageNet: A large-scale hierarchical image database,” IEEE Conference on Computer Vision and Pattern Recognition, USA, pp. 248-255, 2009.
[28]Russakovsky, O., Krause, J., Deng, J., and Berg, A., “ImageNet Large Scale Visual Recognition Challenge (ILSVRC).” Retrieved April, 22, 2018, from www.image-net.org/challenges/LSVRC/, 2010.
[29]Zitzewitz, G. V., “Survey of neural networks in autonomous driving,” ADVANCED SEMINAR (SS), Germany, pp.1-8, 2017.
[30]http://cocodataset.org/#home(2018.5.27).
[31]Lin, T. Y., Maire, M., Belongie, S., Bourdev, L., Girshick, R., Hays, J., Perona, P., Ramanan, D., Zitnick, C. L., and Dollar, P., “Microsoft COCO: Common Objects in Context,” European Conference on Computer Vision (ECCV), pp.740-755, 2014.
[32]Everingham, M., Gool, L., Williams, C. K., Winn, J., and Zisserman, A., "The PASCAL Visual Object Classes (VOC) Challenge," International Journal of Computer Vision, Vol. 88, No. 2, pp.303-338, 2010.
[33]http://explore.digitalglobe.com/spacenet(2018.5.28).
[34]https://www.satellitetoday.com/innovation/2017/01/10/digitalglobes-spacenet-challenge-concludes-first-round-moves-higher-resolution-challenges/(2018.5.28).
[35]https://www.satellitetoday.com/innovation/2017/08/14/digitalglobe-announces-results-spacenet-challenge-round-2/(2018.5.28).
[36]https://www.topcoder.com/spacenet(2018.5.28).
[37]https://medium.com/the-downlinq/object-segmentation-on-spacenet-via-multi-task-network-cascades-mnc-f1c89d790b42(2018.5.28).
[38]http://xviewdataset.org/(2018.5.28).
[39]https://www.wired.com/story/the-pentagon-wants-your-help-analyzing-satellite-images/(2018.5.28).
[40]https://www.iarpa.gov/challenges/fmow.html(2018.5.28).
[41]Christie,G., Fendley, N., Wilson, J., and Mukherjee, R., “Functional Map of the World,” Computer Vision and Pattern Recognition, Cornell University Library, pp.1-16 ,2018.
[42]https://gdo152.llnl.gov/cowc/(2018.5.28).
[43]Mundhenk, T. N., Konjevod, G., Sakla, W. A., and Boakye, K., “A Large Contextual Dataset for Classification, Detection and Counting of Cars with Deep Learning,” European Conference on Computer Vision (ECCV), pp. 785-800, 2016.
[44]Hafemann, L. G., Oliveira, L. S., Cavalin, P. R., and Sabourin, R., ” Transfer learning between texture classification tasks using Convolutional Neural Networks,” International Joint Conference on Neural Networks, Ireland, pp.1-7, 2015.
[45]Pan, S. J. and Yang, Q., “A Survey on Transfer Learning,” IEEE Transactions on Knowledge and Data Engineering, Vol. 22, No. 10, pp.1345-1359, 2010.
[46]Cisek, D., Mahajan, M., Dale, J., Pepper, S., Lin, Y., and Yoo, S., “A transfer learning approach to parking lot classification in aerial imagery,” New York Scientific Data Summit, USA, pp.1-5, 2017.
[47]https://technews.tw/2017/10/05/ai-machine-learning-and-deep-learning/(2018.6.26).
[48]Rosenblatt, F., “The Perceptron : A Probabilistic Model for Information Storage and Organization in the Brain,” Psychological Review, Vol. 65, No. 6, pp.386-408, 1958.
[49]https://zh-tw.coursera.org/courses?query=machine%20learning%20andrew%20ng(2018.5.7).
[50]李宏毅，“一天搞懂深度學習”，台灣資料科學年會，臺北，[DSC 2016]系列活動，2016。
[51]Vinod, S., “INTRODUCTION AND ARTIFICIAL NEURAL NETWORKS,” JERUSALEM COLLEGE OF ENGINEERING, 2017.
[52]https://kknews.cc/zh-tw/news/4mlnr82.html(2018.4.2).
[53]https://www.stockfeel.com.tw/%E6%A9%9F%E5%99%A8%E5%AD%B8%E7%BF%92%E7%9A%84%E8%A1%B0%E9%A0%B9%E8%88%88%E7%9B%9B%EF%BC%9A%E5%BE%9E%E9%A1%9E%E7%A5%9E%E7%B6%93%E7%B6%B2%E8%B7%AF%E5%88%B0%E6%B7%BA%E5%B1%A4%E5%AD%B8%E7%BF%92/(2018.6.26).
[54]https://read01.com/oN2Mkj.html#.WzIE_6cza71(2018.2.5).
[55]Fergus, R., Yu, K., Ranzato, M., Lee, H., Salakhutdinov, R., and Taylor, G., “Deep Learning & Feature Learning Methods for Vision,” Institute for Pure & Applied Mathematics (IPAM), USA, pp.1-67, 2012.
[56]https://medium.com/@yehjames/%E8%B3%87%E6%96%99%E5%88%86%E6%9E%90-%E6%A9%9F%E5%99%A8%E5%AD%B8%E7%BF%92-%E7%AC%AC5-1%E8%AC%9B-%E5%8D%B7%E7%A9%8D%E7%A5%9E%E7%B6%93%E7%B6%B2%E7%B5%A1%E4%BB%8B%E7%B4%B9-convolutional-neural-network-4f8249d65d4f(2018.3.5).
[57]https://www.jeremyjordan.me/batch-normalization/(2018.4.5).
[58]Krizhevsky, A., Sutskever, I., and Hinton, G. E., “ImageNet Classification with Deep Convolutional Neural Networks,” the 25th International Conference on Neural Information Processing Systems, Vol. 1, pp.1097-1105, 2012.
[59]Ioffe, S. and Szegedy, C., “Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift,” Learning, Computer Science, Cornell University Library, pp.1-11, 2015.
[60]Ba, J. L., Kiros, J. R., and Hinton, G. E., “Layer Normalization,” Machine Learning, , Cornell University Library, pp.1-14, 2016.
[61]LeCun, Y., Bottou, L., Bengio, Y., and Haffner, P., “Gradient-based learning applied to document recognition,” Proceedings of the IEEE, Vol.86, No.11, pp.2278-2324, 1998.
[62]Simonyan, K. and Zisserman, A., “Very Deep Convolutional Networks for Large-Scale Image Recognition,” International Conference on Learning Representations, Cornell University Library, pp.1-14, 2015.
[63]Chen, Y. N. and Chang, M., “Convolutional Neural Networks,” 臺灣大學，臺北，pp.1-134, 2016.
[64]zegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., and Wojna, Z., “Rethinking the Inception Architecture for Computer Vision,” IEEE Conference on Computer Vision and Pattern Recognition (CVPR), USA, pp.2818-2826, 2016.
[65]Szegedy, C., Ioffe, S., Vanhoucke, V., and Alemi, A., “Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning,” Computer Vision and Pattern Recognition, Cornell University Library, pp.1-12, 2016.
[66]He, K., Zhang, X., Ren, S., and Sun, J., “Deep Residual Learning for Image Recognition,” IEEE Conference on Computer Vision and Pattern Recognition (CVPR), USA, pp.837-840, 2017.
[67]Huang, G., Liu, Z., Maaen, L.V. D., and Weinberger, K. Q., “Densely Connected Convolutional Networks,” Computer Vision and Pattern Recognition, Computer Science, Cornell University Library, pp.1-9, 2018.
[68]http://funhacks.net/2015/08/12/classifier-evaluation/(2018.5.1).
[69]https://clusteval.sdu.dk/1/clustering_quality_measures/5(2018.5.5).
[70]http://challenge.xviewdataset.org/challenge-description(2018.5.29).
[71]https://kknews.cc/other/9oxr9b.html(2018.5.29).
[72]http://www.pmean.com/definitions/kappa.htm(2018.5.29).
[73]https://medium.com/@14prakash/transfer-learning-using-keras-d804b2e04ef8(2018.5.2).
[74]Li, F.-F., Karpathy, A. and Johnson, J., “Lecture 11：CNNs in Practice,” USA, 2016.
[75]https://www.analyticsvidhya.com/blog/2017/06/transfer-learning-the-art-of-fine-tuning-a-pre-trained-model/(2018.5.24).
[76]Canziani, A., Culurciello, E., and Paszke, A., “An Analysis of Deep Neural Network Models for Practical Applications,” Computer Vision and Pattern Recognition, Cornell University Library, pp.1-7, 2017.
[77]Nakai, E., “Introducton to Convolutional Nerural Network with TensorFlow,” Cloud Solutions Architect at Google, USA, 2017.
[78]陳世儀、張宇洲、蔡季欣、蘇惠璋，“產製網際網路地圖圖磚多重模式之研究”，內政部國土測繪中心，臺北，2014。
[79]Maso, J., Pomakis, K. and Julia, N., “OpenGIS® Web Map Tile Service Implementation Standard,” OpenGIS® Implementation Standard, V.1.0.0, OGC 07-057r7, pp.1-129, 2010.
[80]http://wiki.openstreetmap.org/w/index.php?title=Slippy_map_tilenames&oldid=1552176(2018.5.1)
[81]http://www.maptiler.org/google-maps-coordinates-tile-bounds-projection/(2018.4.28).
[82]http://www.partow.net/miscellaneous/airportdatabase/index.html(2018.4.2).
[83]http://www.poi-factory.com/poifiles/alpha(2017.11.20).