臺灣博碩士論文加值系統

交通事故為全球第八大死亡原因，每年造成近 135 萬人的死傷，根據調查，近五分之一的汽機車交通事故起因為駕駛員自身分心所導致，而現今的駕駛員監控系統(Driver Monitoring System; DMS)對於近幾年發展的自動駕駛車輛來說是必不可少的，該系統能夠警告駕駛員的疲勞以及危險駕駛行為，以便在緊急的情況下接管車輛的控制權。傳統的方法大多為利用機器學習(Machine Learning)方式來擷取特徵並加上分類器實現駕駛分心行為的檢測，而近年來深度學習(Deep Learning)的興起，也開始廣泛地運用在影像處理等應用，受到卷積神經網路(CNN)的影響，愈來愈多基於CNN的駕駛員分心檢測算法被提出，透過輸入靜態影像通過網路提取特徵，最後接上分類器達到分類預測效果，但有些駕駛行為，例如安全駕駛，由於缺乏時間訊息，在靜態圖像中雖然以安全駕駛的行為出現，然而駕駛當下的前後動作可能是其他分心行為(例如從安全駕駛到拿取手機)，在本文中，我們將此問題作為行為識別來處理，除了透過空間的訊息外，還能夠受益於時間的訊息，嘗試提出一種駕駛分心行為的檢測系統，基於雙流卷積融合 (Two-Stream ConvNets) 網路，透過結合空間與時間訊息來檢測出車內檢測駕駛分心行為與原因。

Traffic accidents are the eighth leading cause of death in the world, causing nearly 135 million deaths and injuries every year. According to the report, nearly 1/5 of traffic accidents were caused by the driver’s distractions. The driver monitoring system is essential for the development of self-driving car in recent years, which can warn driver’s dangerous behavior in order to take over control of the car in an emergency. Most of the traditional methods were Machine Learning. Extracting features and add classifiers to detect. In recent years, the popularity of Deep Learning has also begun to be widely used in image processing, and so on. Motivated by the CNN, more and more driver distractions detection and recognition research have been proposed to extract features through the network by input the static image, and finally connected to the categorizer to achieve the classification prediction. In some distracted action, cause the lack of temporal information, the action may be other distractions action. This paper attempts to propose a distraction behavior detection system, based on the Two-Stream ConvNets. By the combination of spatial and temporal information, we can detect the driver’s behavior and reasons of the distractions in the car.

摘要 I
ABSTRACT II
誌謝 III
表目錄 VI
圖目錄 VII
第一章 導論 1
1.1研究動機 1
1.2研究架構 1
第二章 相關文獻 2
2.1 駕駛分心檢測及相關數據 2
2.1.1 Cell Phone Usage Detection 2
2.1.2 UCSD’s Laboratory of Intelligent and Safe Automobiles 3
2.1.3 Southeast University Distracted Driver Dataset 3
2.1.4 StateFarm’s Distracted Driver Detection Dataset 3
2.1.5 AUC Distracted Driver Dataset 4
2.2 影片行為辨識 6
2.1.1 2D ConvNet + LSTM 6
2.1.2 3D ConvNets 7
2.1.3 Two-Stream Networks 8
2.1.4光流(Optical Flow) 8
第三章 系統架構與流程 9
3.1 Two-Stream CNN 9
3.2 訓練流程介紹 11
3.2.1 訓練資料集 11
3.2.2資料輸入前處理 11
3.2.3卷積網路輸入方式 12
第四章 實驗結果 14
4.1實驗環境 14
4.2實驗數據 14
第五章 結論 17
第六章 參考文獻 18

[1]Abouelnaga, Y., Eraqi, H.M., and Moustafa, M.N.: “Real-time distracted driver posture classification”, arXiv preprint arXiv:170609498, 2017
[2]Kaggle https://zh.wikipedia.org/wiki/Kaggle
[3]R. P. A. S. Murtadha D Hssayeni, Sagar Saxena. Distracted driver detection: Deep learning vs handcrafted features. IS&T International Symposium on Electronic Imaging, pages 20–26, 2017
[4]J. Carreira and A. Zisserman, “Quo vadis, action recognition? new models and the kinetics dataset”. In IEEE International Conference on Computer Vision and Pattern Recognition CVPR, 2017
[5]Simonyan, K. and Zisserman, A., “Two-stream convolutional networks for action recognition in videos”, CoRR, abs/1406.2199, 2014. Published in Proc. NIPS, 2014
[6]National highway traffic safety administration traffic safety facts. https://www.nhtsa.gov/risky-driving/distracted-driving/
[7]R. A. Berri, A. G. Silva, R. S. Parpinelli, E. Girardi, and R. Arthur, “A pattern recognition system for detecting use of mobile phones while driving”, In Computer Vision Theory and Applications (VISAPP), 2014 International Conference on, volume 2, pages 411–418. IEEE, 2014.
[8]Y. Artan, O. Bulan, R. P. Loce, and P. Paul, “Driver cell phone usage detection from hov/hot nir images”, In Proceedings of the IEEE conference on computer vision and pattern recognition workshops, pages 225–230, 2014.
[9]C. Craye and F. Karray, “Driver distraction detection and recognition using RGB-D sensor”, arXiv preprint arXiv:1502.00250, 2015.
[10]X. Zhang, N. Zheng, F. Wang, and Y. He, “Visual Recognition of Driver Hand-held Cell Phone Use Based on Hidden CRF”, 2011 IEEE International Conference on Vehicular Electronics and Safety (ICVES), pages 248–251, 2011.
[11]T. Hoang Ngan Le, Y. Zheng, C. Zhu, K. Luu, and M. Savvides, “Multiple Scale Faster-RCNN Approach to Driver’s Cell-Phone Usage and Hands on Steering Wheel Detection”, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, pages 46–53, 2016.
[12]K. Seshadri, F. Juefei-xu, D. K. Pal, M. Savvides, and C. P. Thor, “Driver Cell Phone Usage Detection on Strategic Highway Research Program ( SHRP2 ) Face View Videos”, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, pages 35—- 43, 2015.
[13]S. Martin, E. Ohn-Bar, A. Tawari, and M. M. Trivedi, “Understanding head and hand activities and coordination in naturalistic driving videos”, In Intelligent Vehicles Symposium Proceedings, 2014 IEEE, pages 884–889. IEEE, 2014.
[14]E. Ohn-bar and S. Martin, “Driver hand activity analysis in naturalistic driving studies : challenges , algorithms , and experimental studies challenges , algorithms , and experimental studies”, Journal of Electronic Imaging, 22(4), 2013.
[15]C. H. Zhao, B. L. Zhang, J. He, and J. Lian, “Recognition of driving postures by contourlet transform and random forests”, IET Intelligent Transport Systems, 6(2):161–168, 2011.
[16]StateFarm’s Distracted Driver Detection Dataset https://www.kaggle.com/c/state-farm-distracted-driver-detection
[17]K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition”, International Conference on Learning Representations.
[18]B. Baheti, S. Gajre, and S. Talbar, “Detection of distracted driver using convolutional neural network,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, pp. 1032–1038.
[19]Hesham M. Eraqi, Yehya Abouelnaga, Mohamed H. Saad, and Mohamed N. Moustafa, “Driver Distraction Identification with an Ensemble of Convolutional Neural Networks”, Journal of Advanced Transportation, Machine Learning in Transportation (MLT) Issue, 2019.
[20]Khurram Soomro, Amir Roshan Zamir and Mubarak Shah, “UCF101: A Dataset of 101 Human Action Classes From Videos in The Wild”, CRCV-TR-12-01, November, 2012.
[21]I. Laptev and T. Lindeberg, “Space-time interest points”, In Proc. ICCV, pages 432–439, 2003
[22]P. Dollar, V. Rabaud, G. Cottrell, and S. Belongie, “Behavior recognition via sparse spatio-temporal features”, In Proc. ICCV VS-PETS, 2005.
[23]S. Sadanand and J. Corso, “Action bank: A high-level representation of activity in video”, In CVPR, 2012.
[24]H. Wang and C. Schmid, “Action recognition with improved trajectories”, In ICCV, 2013.
[25]World Health Organization Global status report on road safety 2018 https://www.nhtsa.gov/risky-driving/ distracted-driving/
[26]Nikhil Das, Eshed Ohn-bar, and Mohan M Trivedi, “On Performance Evaluation of Driver Hand Detection Algorithms : Challenges , Dataset , and Metrics.” 2015.