進階搜尋


   電子論文尚未授權公開,紙本請查館藏目錄
(※如查詢不到或館藏狀況顯示「閉架不公開」,表示該本論文不在書庫,無法取用。)
系統識別號 U0026-1507201614355200
論文名稱(中文) 基於電腦視覺技術之汽車車門安全裝置設計與實現
論文名稱(英文) Design and Implementation of Car Safety Device based on Computer Vision Technique
校院名稱 成功大學
系所名稱(中) 工程科學系
系所名稱(英) Department of Engineering Science
學年度 104
學期 2
出版年 105
研究生(中文) 葉宥成
研究生(英文) You-Cheng Yeh
學號 N96031033
學位類別 碩士
語文別 英文
論文頁數 48頁
口試委員 指導教授-廖德祿
口試委員-顏錦柱
口試委員-簡尊彝
中文關鍵字 影像辨識  機器學習  哈爾特徵  影像光流  分類器  被動式安全裝置 
英文關鍵字 Image recognition  machine learning  Haar features  optical flow  classifier  safety device. 
學科別分類
中文摘要 近年來,在嵌入式電腦、感測器與網通的效能增長下,車用電子被各大公司視為一塊藍海市場,從胎壓偵測器、行車紀錄器、預防碰撞系統、環景影像系統到近年來的無人車,越來越多相關應用被開發出來,尤其智慧型手機成長漸漸飽和的狀況下,企業為了尋求利潤,漸漸的開始投資車用電子這個部分,而車用電子最注重於安全性,因此,車用式安全裝置成為研究重點之一。
本論文中所開發之被動式安全裝置,主要應用於汽車路邊停車時,警示後方有無機車、腳踏車…等交通工具即將行駛過汽車旁邊,以避免汽車駕駛或乘客路邊停車時,因在下車前並未向後察看有無來車就直接開門,而造成之交通意外。
本論文將主要以電腦視覺演算法中的哈爾特徵(Haar Features)搭配自適性增強(Adaptive Boosting)演算法共同訓練分類器,接著,利用影像光流 (Lucas-Kanade Optical Flow) 演算法過濾錯誤資訊,最終,實作在嵌入式電腦上,實作上須考慮影像處理之運算速度,以達到實時(Real-Time)的效果,因此選用核心(CPU)工作時脈為900MHz並搭載GPU之樹梅Pi2(Raspberry Pi2)嵌入式電腦,而程式碼部分,則使用影像處理函式庫OpenCV撰寫。
本論文會詳細解釋使用的電腦視覺演算法之原理,透過理解原理並結合現實生活中可能的情況,來發展適合機車辨識的使用方式,最後會統計辨識正確率與錯誤告警率的成果,同時,也會一併討論此成果之問題點。
英文摘要 In recent years, due to the progress of embedded computer, sensor, and network, the automotive electronic market starts to be considered as a blue ocean market. More and more applications about automotive electronic that such as TPMS, driving records, pre-crash system, AVM and self-driving car were developed, and since the smart phone market reached saturation, corporations eager to find out more ways to make profit, they gradually begin investing more resource in this market. However, the security is most important part of automotive electronic, and that is why “safety device” becomes one of the key directions in research field.
This thesis mainly aims to avoid accident that caused by driver or passenger who forget to check if any motorcycle or bicycle roared was passing by or getting near to them while getting off.
The classifier of this thesis was trained by Haar features with adaptive boosting and further filter the wrong target via Lucas-Kanade optical flow algorithm. Finally, it was implemented through embedded system. In order to consider the computing rate (for Real Time), choose “Raspberry Pi2” to implement, because this embedded computer has 900Mhz CPU clock and GPU. In addition, the program was developed via OpenCV library.
This thesis will carefully explain theory about computer algorithms that are used, through comprehending those theories and probable situations in reality. It could be easier to develop better methods for “motorcycle recognition”. In the end, this thesis will count out the correct rate and false alarm rate, in the meantime, problems will be concluded.
論文目次 摘要 I
ABSTRACT II
誌謝 IV
CONTENTS V
LIST OF TABLE VII
LIST OF FIGURE VIII
CHAPTER 1 INTRODUCTION 1
1.1 BACKGROUND 1
1.2 MOTIVATION AND OBJECTIVES 3
1.3 THESIS ORGANIZATION 4
CHAPTER 2 FUNDAMENTAL BACKGROUND 5
2.1 IMAGE PRE-PROCESSING 5
2.1.1 CONVERT TO GRAY IMAGE 5
2.1.2 HISTOGRAM EQUALIZATION 6
2.2 HAAR-LIKE FEATURES 8
2.2.1 RECTANGULAR HAAR-LIKE FEATURES 8
2.2.2 INTEGRAL GRAPH 9
2.3 ADABOOST ALGORITHM 10
2.3.1 MATHEMATICAL FORMUALS OF ADABOOST ALGORITHM 10
2.3.2 ADABOOST TRAINING RECIPE 11
2.4 SYMMETRICAL MATCHING 15
2.5 LUCAS-KANADE OPTICAL FLOW 16
2.6 INTRODUCTION OF RASPBERRY PI2 18
CHAPTER 3 SYSTEM DESIGN 20
3.1 SYSTEM STRUCTURE 20
3.2 TRAINING SYSTEM 22
3.2.1 TRAINING SAMPLES COLLECTING 22
3.2.2 HAAR-LIKE FEATURE PICKING 23
3.2.3 ADABOOST TRAINING METHOD 25
3.3 RECOGNITION SYSTEM 26
3.3.1 HAAR FEATURES IDENTIFICATION 27
3.3.2 REMOVE THE SUBSTANDARE AFTER SYMMETRICAL MATCHING 28
3.3.3 HAZARD ESTIMATION 29
3.4 ALARM SYSTEM 30
CHAPTER 4 TESTING AND VERIFICATION 33
4.1 THE TEST PARAMETERS 33
4.2 ACCURACY TEST BASED ON DEPTH 34
4.3 ALARM RATE TEST BASED ON SPEED 40
CHAPTER 5 CONCLUSIONS AND FUTURE WORK 45
5.1 CONCLUSION 45
5.2 FUTURE WORK 46
REFERENCE 47
參考文獻 [1] Pollard, Stephen B., John EW Mayhew, and John P. Frisby. "PMF: A stereo correspondence algorithm using a disparity gradient limit." Perception 14.4 (1985): 449-470.
[2] Marola, Giovanni. "Using symmetry for detecting and locating objects in a picture." Computer Vision, Graphics, and Image Processing, pp. 179-195, 1989.
[3] Freund, Yoav, Robert Schapire, and N. Abe. "A short introduction to boosting." Journal-Japanese Society for Artificial Intelligence 14.771-780, Sept. 1999.
[4] Klotz, Michael, and Hermann Rohling. "24 GHz radar sensors for automotive applications." Microwaves, Radar and Wireless Communications. 2000. MIKON-2000. 13th International Conference on. Vol. 1. IEEE, 2000.
[5] Viola, Paul, and Michael Jones. "Rapid object detection using a boosted cascade of simple features." Computer Vision and Pattern Recognition, 2001. CVPR 2001. Proceedings of the 2001 IEEE Computer Society Conference on. Vol. 1. IEEE, 2001.
[6] Bouguet, Jean-Yves. "Pyramidal implementation of the affine lucas kanade feature tracker description of the algorithm." Intel Corporation 5.1-10 (2001): 4.
[7] R. Lienhart, J. Maydt. "An Extended Set of Haar-like Features for Rapid Object Detection", ICIP 2002, vol.1, pp. 900-903, 2002.
[8] Demirkir, Cem, and B. Sankur. "Face detection using boosted tree classifier stages.", Signal Processing and Communications Applications Conference, pp. 575-578, April. 2004.
[9] Y. Freund, R. Schapire(2004). A Tutorial on Boosting[Online]. Avaiilable:http://www.di.unipi.it/~cardillo/AA0304/fabio/boosting.pdf (accessed 12 Oct. 2015), 2004.
[10] T. Mita, T. Kaneko, O. Hori. "Joint Haar-like features for face detection", ICCV 2005, vol.2, pp. 1619-1626, Oct. 2005.
[11] Khammari, Ayoub, et al., "Vehicle detection combining gradient analysis and AdaBoost classification.", Intelligent Transportation Systems, 2005. Proceedings. 2005 IEEE., pp. 66-71, Sept. 2005.
[12] Abdullah-Al-Wadud, Mohammad, et al. "A dynamic histogram equalization for image contrast enhancement.", IEEE Trans. on Consumer Electronics, pp. 593-600, July. 2007.
[13] Goffredo, Michela, John N. Carter, Mark S. Nixon., "Front-view gait recognition.", Biometrics: Theory, Applications and Systems, 2008. BTAS 2008. 2nd IEEE International Conference on., IEEE, 2008.
[14] Wu, Yan-Wen, and Xue-Yi Ai. "Face detection in color images using AdaBoost algorithm based on skin color information." Knowledge Discovery and Data Mining, 2008. WKDD 2008. First International Workshop on., pp. 339-342, 2008.
[15] Chen, C. T., and Y. S. Chen. "Real-time approaching vehicle detection in blind-spot area." 2009 12th International IEEE Conference on Intelligent Transportation Systems. IEEE, 2009.
[16] M. S. S. Kumar, K. S. Vimala,N. Avinash., "Face distance estimation from a monocular camera", IEEE International Conference on Image Processing(ICIP)., pp. 3532-3536, Sept. 2013.
[17] Leen, Gabriel, Donal Heffernan. (2002 Jan. 07). Expanding automotive electronic systems. Computer Vol.35[Online]. pp. 88-93. Available:http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=976923
[18] Choi, Jaesik. Realtime On-Road Vehicle Detection with Optical Flows and Haar-Like Feature Detectors[Online]., Available:https://www.ideals.illinois.edu/handle/2142/30794 (accessed 22 Nov. 2015), 2012.
[19] OpenCV Documentation,
Available:http://docs.opencv.org/2.4/modules/refman.html
[20] OpenCV Documentation, Anailable:http://docs.opencv.org/2.4/doc/user_guide/ug_traincascade.html
[21] Bradski, Gary, Adrian Kaehler., Learning OpenCV: Computer vision with the OpenCV library., 2nd ed, O'Reilly Media, Inc., 2008.
[22] Brahmbhatt, Samarth., Embedded Computer Vision: Running OpenCV Programs on the Raspberry Pi. Practical OpenCV., Apress, 2013. 201-218
[23] 毛星云., OpenCV3 編程入門., 電子工業出版社., 2015
論文全文使用權限
  • 同意授權校內瀏覽/列印電子全文服務,於2021-07-08起公開。
  • 同意授權校外瀏覽/列印電子全文服務,於2021-07-08起公開。


  • 如您有疑問,請聯絡圖書館
    聯絡電話:(06)2757575#65773
    聯絡E-mail:etds@email.ncku.edu.tw