基於深度學習之公車人數計算｜國立清華大學博碩士論文庫

簡易檢索 / 詳目顯示

回結果列表

研究生：	邱毅思 Anyony Muthu, Anto Joeis
論文名稱：	基於深度學習之公車人數計算 Smart Public Transport Using Deep Learning Method
指導教授：	黃能富 Huang, Nen-Fu
口試委員:	廖冠雄陳懷恩
學位類別：	碩士 Master
系所名稱：	電機資訊學院 - 通訊工程研究所 Communications Engineering
論文出版年：	2020
畢業學年度：	108
語文別：	英文
論文頁數：	41
中文關鍵詞：	深度學習、卷積神經網路、長波紅外線相機、樹莓派、物體偵測
外文關鍵詞：	Deep learning, Convolutional Neural Network (CNN), Long-Wave Infrared (LWIR) Camera, Raspberry Pi, Object Detection
相關次數：	點閱：2 下載：0
分享至:	分享至facebook 分享至twitter

查詢本校圖書館目錄查詢臺灣博碩士論文知識加值系統勘誤回報

人工智慧是經由利用電腦計算使機器呈現出人類的智慧的技術。人工智慧的近期研究發展也讓不同的領域像是大眾交通方面，迎來了許多新的契機。大眾交通是全球各地民眾日常生活中的必不可少的服務。儘管大眾運輸業因不同類型的技術不斷提高服務水準，在此篇論文採取深度學習技術去提供車上可用座位訊息。我著重於偵測小型巴士中的乘客數量，因此推斷出可用的座位。第一步，在校園公車上使用Lepton-Long Wave紅外攝影機收集大量的紅外線攝像，並使用影像處理技術和資料擴充(data augmentation )方法處理圖像資料，這樣的方式不但解決白天和晚上的環境中的辨識問題，並且也不會觸犯到隱私。處理好的圖像資訊結合兩種不同的深度學習模型Faster R-CNN和SSD，最後選擇在樹梅派主機上運行較佳的評估模型。最後將此人工智慧轉移到樹莓派主機，並結合樹莓派紅外線攝影機，以高準確度檢測出校園公車上的人數。

Artificial intelligence (AI) is the simulation of human intelligence, which is processed by machines or computer systems. The studies and research were done on the Artificial Intelligence (AI) in the recent past have opened new doors of opportunities to different sectors including the public transport sector. Public transport is an essential service embedded in day to day life of common people worldwide. Though the transport sector has been constantly upgraded by different kind of technologies, here we tried deep learning method to provide information about available seats in real-time. In this thesis, I focus on detecting the number of passengers in a minibus. Consequently, the available empty seats will be known. Initially, I have collected the Infrared image using Lepton-Long Wave Infrared Camera for detecting in both day and night environment and also overcome privacy issue. I have followed the image processing and data augmentation method to modify the image data and evaluate ththe modifieddata using two different deep learning models Faster R-CNN and SSD. Choose different models for analysis performance suitable for Raspberry Pi. The trained model is transfer to Raspberry Pi and using Infrared Camera detection. The number of people presents inside the bus was detected successfully with a high accuracy rate.

ABSTRACT .....................................................................................i
摘要 ................................................................................................ii
TABLE OF CONTENTS ...................................................................iii
LIST OF FIGURES ...........................................................................v
LIST OF TABLES ............................................................................vii
CHAPTER 1 .....................................................................................1
INTRODUCTION  .............................................................................1
CHAPTER 2 .....................................................................................4
BACKGROUND AND RELATED WORKS ...........................................4
2.1 Thermal Imaging  .......................................................................4
2.2 Deep Learning  ..........................................................................5
2.3 Convolutional Neural Network  ..................................................7
2.4 Tensorflow Object Detection API  .............................................9
2.4.1 Faster R-CNN  .......................................................................10
2.4.2 Single Shot Multi-Box Detector (SSD)  .................................12
2.5 Artificial Intelligence Used in Public Transportation  ...............13
CHAPTER 3 ....................................................................................16
System Design  ..............................................................................16
3.1 Hardware Device  .....................................................................17
3.1.1 Camera Module  .....................................................................17
3.1.2 System  ..................................................................................18
3.2 Data Construction  ...................................................................19
3.2.1 Data Collection  .....................................................................19
3.2.2 Data Processing  ...................................................................19
3.3 Tensorflow Training  ................................................................20
CHAPTER 4  ...................................................................................21
System Implementation  .................................................................21
4.1 Data Collection ..........................................................................21
4.2 Data Processing ........................................................................23
4.2.1 Data Extraction .......................................................................23
4.2.2 Image Segmentation ..............................................................24
4.2.3 Image Annotation ...................................................................26
4.3 Training Model Architecture ......................................................27
4.3.1 CNN Training Model  ...............................................................27
4.3.2 SDD Training Model ................................................................30
CHAPTER 5 ......................................................................................31
Experimental Results  .......................................................................31
5.1 Tensorflow Object detection API with GPU .................................31
5.1.1 Faster R-CNN Training and Testing ...........................................31
5.1.2 SSD Training and Testing ..........................................................34
5.1.3 Comparison of Faster RCNN and SDD. .....................................35
5.2 Raspberry Pi Result. ....................................................................36
CHAPTER 6 ........................................................................................37
Conclusion and Future Work ..............................................................37
6.1 Conclusion ....................................................................................37
6.2 Future Work ..................................................................................38
References ..........................................................................................39
                                

1. Urban public transport in 21st century, www.utip.org
2. Militaru, I. G. D. Deep learning in acoustic modeling for Automatic Speech Recognition and Understanding - an overview. International Conference on Speech Technology and Human-Computer Dialogue (SpeD) (2015).
3. Md Tohidul Islam, B. M. N. K. S., Sagidur Rahman, Taskeed Jabid. Image Recognition with Deep Learning. International Conference on Intelligent Informatics and Biomedical Sciences (ICIIBMS) (2018).
4. Tom Young, D. H., Soujanya Poria, Erik Cambria. Recent Trends in Deep Learning Based Natural Language Processing. arXiv:1708.02709v8 (2018).
5. Zubair Md. Fadlullah, F. T., Bomin Mao, Nei Kato. State of the Art Deep Learning: Evolving Machine Intelligence Toward Tomorrow’s Intelligent Network Traffic Control Systems. IEEE Communications Surveys Tutorials VOL. 19, NO. 4 (2017).
6. Wang Yongqing, G. Z., Wang Shuonan, He Ping. The temperature measurement technology of infrared thermal imaging and its applications review. IEEE International Conference on Electronic Measurement & Instruments (2017).
7. Y.leChun, Y. B., and G.Hilton. Deep learning. Nature vol. 521, no.7553, p.436 (2015).
8. K. He, X. Z., S. Ren, and J. Sun. Deep residual learning for image recognition. arXiv:1512.03385 (2015).
9. Tianyi Liu, S. F., Yuehui Zhao, Peng Wang. Implementation of Training Convolutional Neural Networks. ArXiv, 1506.01195 (2015).
10. Kaiming He, X. Z., Shaoqing Ren, Jian Sun. Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification. The IEEE International Conference on Computer Vision (ICCV) pp. 1026-1034 (2015).
11. Jonathan Huang, V. R., Chen Sun, Menglong Zhu, Anoop Korattikara, Alireza Fathi, Ian Fischer, Zbigniew Wojna, Yang Song, Sergio Guadarrama, Kevin Murphy. Speed, accuracy trade-offs for modern convolutional object detectors. arXiv:1611.10012 (2016).
12. Kaiming He, G. G., Piotr Dollár, Ross Girshick. Mask R-CNN. arXiv:1703.06870 (2018).
13. Tsang, S.-H. Review: SSD — Single Shot Detector (Object Detection). www.towarddatascience.com.
14. Shaoqing Ren, K. H., Ross Girshick, Jian Sun. Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks. arXiv:1506.01497 (2015).
15. Jifeng Dai, Y. L., Kaiming He, Jian Sun. R-FCN: Object Detection via Region-based Fully Convolutional Networks. Advances in Neural Information Processing Systems 29 (NIPS) (2016).
16. M. Everingham, L. V. G., C. K. Williams, J. Winn, A. Zisserman. The pascal visual object classes (VOC). International Journal of Computer Vision (2010).
17. Zitnick, T.-Y. L. M. B. H. P. R. D. L. Microsoft COCO: Common Objects in Context. European Conference on Computer Vision, pp 740-755 (2014).
18. Méndez-Vázquez, R. D. F. S. M. Face Recognition Using TOF, LBP and SVM in Thermal Infrared Images. Iberoamerican Congress on Pattern Recognition, pp 683-691 (2011).
19. Wei Liu, D. A., Dumitru Erhan, Christian Szegedy, Scott Reed, Cheng-Yang Fu, Alexander C. Berg. SSD: Single Shot Multi-Box Detector. arXiv:1512.02325, 21-37 (2016).
20. Zisserman, K. S. a. A. Very deep convolutional networks for large-scale image recognition. arXiv:1409.1556 (2014).
21. Parks, K. Purethermal1-uvc-capture. www.github.com.
22. Mike Folk, E. P. Balancing Performance and Preservation Lessons learned with HDF5. US DPIF Workshop (2010).
23. Yigit Baran Can, R. T. An efficient CNN for spectral reconstruction from RGB images. arXiv:1804.04647 (2018).
24. Qianhui Luo, H. M., Yue Wang, Li Tang, Rong Xiong. 3D-SSD: Learning Hierarchical Features from RGB-D Images for Amodal 3D Object Detection arXiv:1711.00238v2 (2018).
25. Christian Szegedy, V. V., Sergey Ioffe, Jon Shlens, Zbigniew Wojna. Rethinking the Inception Architecture for Computer Vision. arXiv:1512.00567 (2015).
26. Howard AG, Z. M., Chen B, Kalenichenko D, Wang W, Weyand T, Andreetto M, Adam H, MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications. arXiv:1704.04861 (2017).
27. Faizan Shaikh, Deep Learning vs. Machine Learning, www.analyticsvidhya.com.
28. Zheng Zhaoping, Zeng Hansheng, Ding Cuijiao, et al. Summary on the infrared thermal Imaging temperature measurement technology and its application. Infrared Technology, 2003; (01):96-98.
29. Sarah Wray. HERE and Verizon partner for 5G-enabled safety and navigation. www.smartcitiesworld.net.

簡易檢索 / 詳目顯示

相關論文