本論文開發了具深度學習的嵌入式車道線偵測系統。此系統具有偵測結構化與非結構化道路之能力，可適應於多種類的場景，並給予穩定的車道線資訊。本論文透過卷積自編碼(Convolutional Auto-Encode，CAE)神經網路的降噪以及圖像重構之特性，將圖像中車道線以外的所有物件濾除，進而提取車道線之特徵，並以雙曲線模型擬合車道線。最後，利用粒子濾波演算法進行車道線追蹤，且實現於NVIDIA Jetson TX2平台上。
本論文為了驗證演算法效能，測試場景分為日間、夜間以及雨天等三種不同的天候。除了一般的結構化道路外，並包含其他特殊場景，如陰影、隧道、非結構、車道線退化磨損、以及車道線遭車輛遮擋等。經由驗證的數據顯示，本論文提出的結構化與非結構化道路之車道線偵測系統的正確率為90.06 %。

An embedded lane detection system based on deep learning is presented in this article. The system has the ability to detect structured and unstructured roads. Stable information on lane markings in a variety of scenarios is also provided by the system. In order to get a clearer image with lane markings, a Convolution Auto-Encode (CAE) that has the characteristics of noise reduction and reconstruction is used to remove all the objects in the images except lane markings. By lane feature extraction and hyperbolic model, the feature points of lane markings are fitted. Finally, the particle filter is used for lane tracking, and the whole system is implemented on the NVIDIA Jetson TX2 platform.
At last, three different situations, such as daytime, night, and rainy day, are selected to demonstrate the performance of the proposed algorithm. In addition to deal with general structured roads, some special scenes, such as shadow, tunnel, unstructured roads, degenerate lane markings, and blocked lane markings, are also considered. According to the experiment result, the accuracy of the proposed lane detection system for structured and unstructured roads is 90.06 %.

目錄
論文審定書 i
致　謝 ii
摘　要 iii
Abstract iv
目 錄 v
圖　次 vii
表　次 xi
第一章 緒 論 1
1-1 研究動機 1
1-2 文獻回顧 2
1-3 主要貢獻 9
1-4 章節介紹 10
第二章 系統概述 11
2-1 系統平台 11
2-1-1 硬體設備與作業系統 11
2-1-2 軟體開發套件 12
2-2 系統架構流程 14
第三章 系統實現 16
3-1 車道線提取 16
3-1-1 卷積神經網路 16
3-1-2 自編碼神經網路 25
3-1-3 卷積自編碼神經網路 26
3-2 車道雙曲線擬合 30
3-2-1 RANSAC演算法 31
3-2-2 特徵點搜尋算法 32
3-2-3 多項式曲線擬合 34
3-2-4 多消失點偵測 35
3-2-5 M-估計法 38
3-2-6 雙曲線建構 40
3-3 車道線追蹤 42
3-3-1 貝氏濾波器 42
3-3-2 蒙地卡羅方法 43
3-3-3 粒子濾波演算法 44
第四章 實驗結果 49
4-1 實驗場景介紹 49
4-2 車道線提取 51
4-3 車道線偵測結果 55
第五章 結論與未來展望 68
5-1 結論 68
5-2 未來展望 68
參考文獻 69

[1] Taiwan Area National Freeway Bureau，MOTC。
取自：https://www.freeway.gov.tw/Publish.aspx?cnid=516&p=128
[2] J. Han, D. Kim, M. Lee, and M. Sunwoo, “Road boundary detection and tracking for structured and unstructured roads using a 2D lidar sensor,” International Journal of Automotive Technology, vol. 15, pp. 611-623, 2014.
[3] S. C. Yi, Y. C. Chen, and C. H. Chang, “A Lane Detection Approach Based on Intelligent Vision,” Computers and Electrical Engineering, vol. 42, pp. 23-29, 2015.
[4] N. Madrid, and P. Hurtik, “Lane departure warning for mobile devices based on a fuzzy representation of images,” Fuzzy Sets and Systems, vol. 92, pp. 319-334, 2015.
[5] J. Niu, J. Lu, M. Xu, P. Lv, and X. Zhao, “Robust Lane Detection using Two-stage Feature Extraction with Curve Fitting,” Pattern Recognition, vol. 59, pp. 225-233, 2016.
[6] A. Mammeri, A. Boukerche, and Z. Tang, “A real-time lane marking localization tracking and communication system,” Computer Communications, vol. 73, pp. 132-143, 2016.
[7] K. Zhaom, M. Meuter, C. Nunn, D. Muller, S. M. Schneiders, and J. Pauli, “A novel multi-lane detection and tracking system,” IEEE Conference on Intelligent Vehicles Symposium (IV), pp.1084-1089, 2012.
[8] Y. Wang, N. Dahnoun, and A. Achim, “A novel system for robust lane detection and tracking,” Signal Process, vol. 92, pp. 319-334, 2012.
[9] L. Bai, Y. Wang, and M. Fairhurst, “An extended hyperbola model for road tracking for video-based personal navigation,” Knowledge-Based Systems, vol. 21, pp. 265-272, 2008.
[10] L. Bai, Y. Wang, and M. Fairhurst, “Robust Road Modeling and Tracking Using Condensation,” IEEE Transactions on Intelligent Transportation Systems, vol. 9, pp. 570-579, 2008.
[11] L. Bai, Y. Wang, and M. Fairhurst, “Multiple condensation filters for road detection and tracking,” Pattern Analysis and Applications, vol. 13, pp. 251-262, 2010.
[12] L. Bai, and Y. Wang, “Road tracking using particle filters with partition sampling and auxiliary variables,” Computer Vision and Image Understanding, vol. 115, pp. 1463-1471, 2011.
[13] 王海、蔡英鳳、林國餘、張為公，基於方向可變Haar特徵和雙曲線模型的車道線檢測方法，交通運輸工程學報，2014。
[14] W. Ding, and Y. Li, “Efficient vanishing point detection method in complex urban road environments,” IET Computer Vision, vol. 9, pp. 549-558, 2015.
[15] W. Ding, Y. Li, and H. Liu, “Efficient vanishing point detection method in unstructured road environments based on dark channel prior,” IET Computer Vision, vol. 10, pp. 852-860, 2016.
[16] S. L. Phung, M. C. Le, and A. Bouzerdoum, “Pedestrian lane detection in unstructured scenes for assistive navigation,” Computer Vision and Image Understanding, vol. 149, pp. 186-196, 2016.
[17] X. Fan, and H. Shin, “Road vanishing point detection using weber adaptive local filter and salient-block-wise weighted soft voting,” IET Computer Vision, vol. 10, pp. 503-512, 2016.
[18] J. Shi, J. Wang, and F. Fu, “Fast and Robust Vanishing Point Detection for Unstructured Road Following,” IEEE Transactions on Intelligent Transportation Systems, vol. 17, pp. 970-979, 2015.
[19] H. Y. Cheng, C. C. Yu, C. C. Tseng, K. C. Fan, J. N. Hwang, and B. S. Jeng, “Environment classification and hierarchical lane detection for structured and unstructured roads,” IET Computer Vision, vol. 4, pp. 37-49, 2009.
[20] J. Kim, J. Kim, G. J. Jang, and M. Lee, “Fast learning method for convolutional neural networks using extreme learning machine and its application to lane detection,” Neural Networks, vol. 87, pp. 109-121, 2017.
[21] Y. Tian, J. Gelernter, X. Wang, W. Chen, J. Gao, Y. Zhang, and X. Li, “Lane marking detection via deep convolutional neural network,” Neurocomputing, vol. 280, pp. 46-55, 2018.
[22] D. Neven, B. D. Brabandere, S. Georgoulis, M. Proesmans, and L. V. Gool, “Towards End-to-End Lane Detection: an Instance Segmentation Approach,” arXiv preprint arXiv:1802.05591, 2018.
[23] X. Pan, J. Shi, P. Luo, X. Wang, and X. Tang, “Spatial as deep: Spatial cnn for traffic scene understanding,” In AAAI Conference on Artificial Intelligence (AAAI), 2018.
[24] NVIDIA，GEFORCE GTX 1080 Ti。
取自: https://www.geforce.com.tw/graphics-cards/geforce/pascal/tw/gtx-1080-ti/
[25] NVIDIA Jetson TX2。取自:
http://www.nvidia.com.tw/object/embedded-systems-dev-kits-modules-tw.html
[26] NVIDIA，嵌入式系統。取自:
http://www.nvidia.com.tw/object/embedded-systems-dev-kits-modules-tw.html
[27] 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，2017。
[28] 林大貴，Keras深度學習簡介。
取自: http://tensorflowkeras.blogspot.tw/2017/08/keras.html
[29] TensorFlow中文社區。取自: http://www.tensorfly.cn/
[30] 使用Tensorflow實作卷積神經網路。
取自：http://arbu00.blogspot.tw/2017/03/2-tensorflowconvolutional-neural.html
[31] N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhutdinov, “Dropout: A Simple Way to Prevent Neural Networks from Overfitting,” Journal of Machine Learning Research, vol. 15, pp. 1929-1958, 2014.
[32] 激勵函數-Activation Function。
取自：https://blog.csdn.net/dabokele/article/details/58713727
[33] M. D. Zeiler, “ADADELTA: An Adaptive Learning Rate Method,” arXiv preprint arXiv:1212.5701, 2012.
[34] D. Kingma, and J. Ba, “Adam: A method for stochastic optimization,” International Conference on Learning Representations (ICLR), pp. 1-15, 2015.
[35] lan Goodfellow、Yoshua Bengio、Aaron Courville，深度學習Deep Learning，人民郵電出版社，2017。
[36] C. Y. Liou, W. C. Cheng, J. W.Liou, and D. R. Liou, “Autoencoder for words,” Neurocomputing, vol. 138, pp. 84-96, 2014.
[37] V. Dumoulin, and F. Visin, “A guide to convolution arithmetic for deep learning,” arXiv preprint arXiv:1603.07285, 2016.
[38] M. A. Fischler, and R. C. Bolles, “Random Sample Consensus: A Paradigm for Model Fitting with Applications to Image Analysis and Automated Cartography,” Readings in Computer Vision, vol. 24, pp. 726-740, 1987.
[39] Y. U. Yim, and S. Y. Oh, “Three-feature based automatic lane detection algorithm (TFALDA) for autonomous driving,” IEEE Transactions on Intelligent Transportation Systems, vol. 4, pp. 219-225, 2003.
[40] 張維仁，以DSP為基礎實現道路偵測與車輛控制，國立中山大學機械與機電工程學系，碩士論文，2011。
[41] D. L. Massart, L. Kaufman, P. J. Rousseeuw, and A. Leroy, “Least median of squares: a robust method for outlier and model error detection in regression and calibration,” Analytica Chimica Acta, vol. 187, pp. 171-179, 1986.
[42] RANSAC LMedS 詳細分析。
取自: https://blog.csdn.net/billbliss/article/details/78592216
[43] P. C. McLachlan, “On the generalised distance in statistics,” Proceedings of the National Institute of Sciences of India, vol. 2, pp. 49-55, 1936.
[44] S. Xiang, F. Nie, and C. Zhang, “Learning a Mahalanobis distance metric for data clustering and classification,” Pattern Recognition, vol. 41, pp. 3600-3612, 2008.
[45] P. J. Huber, “Robust Estimation of a Location Parameter,” The Annals of Mathematical Statistics, vol. 35, pp. 73-101, 1964.
[46] E. Orhan, “Particle Filtering,” Center for Neural Science of New York University Sciences, 2012.
[47] 全國法規資料庫。
取自：http://law.moj.gov.tw/LawClass/LawAll.aspx?PCode=K0040014
[48] 鍾凱全，以雷達與影像融合技術車前防碰撞警示系統，國立中山大學機械與機電工程學系，碩士論文，2017。
[49] 陳信志，基於粒子濾波器之都市環境車道偵測，國立高雄第一科技大學電腦與通訊工程系，碩士論文，2015。