車用電子一直是近年來相當火熱的市場，自動駕駛更是各大車商積極開發的領域，而這不外乎先進駕駛輔助系統的迅速成長與發展。本論文看準行車紀錄器的普及，以單鏡頭的攝影機開發前方碰撞預警系統，並設置偵測範圍50公尺以內，以達到實時性、穩定性與強健性為目標。
演算法的部分分為串聯式分類器的訓練與三大區塊流程，車輛偵測、車輛追蹤與碰撞判斷。我們利用哈爾特徵與自適應增強演算法的結合，使用Active-learning Framework訓練一個好的串聯式分類器。車輛偵測則是使用兩階段的架構，第一階段，串聯式分類器用多尺度收尋視窗找尋車輛候選物件，第二階段，進一步篩選降低錯誤偵測，並將確認是車輛的物件，取得更好的車輛寬資訊，以便最後碰撞判斷有更穩定的效果。接著在相鄰畫面辨識車輛，達到追蹤。最後在有可能發生碰撞的區塊中找尋優先權最高的車輛，並利用車寬資訊與取樣時間計算碰撞時間，如果小於兩秒則發出警告提示駕駛者。
接著我們分析軟體演算法各個區塊的運行時間，第一階段的車輛偵測耗費大部分的時間，也是系統最前端的區塊，故本論文針對此部分設計硬體達到加速運算，硬體設計架構方面，包含了切割與縮放、積分影像產生、line buffers、收尋視窗暫存器與平行化之串聯式分類器。將畫面切割與縮放可以大幅的減少硬體資源的使用，採用兩個收尋視窗暫存器直接減少了整體處理時間的一半，也可以讓line buffers的資料有更高的重複使用性，平行化之串聯式分類器直接將串聯式的階層打平，快速的判斷出是否為車輛，來達到我們的目標60 FPS。

Automotive electronics has been a very hot market in recent years, autopilot is the major field to many car companies, which is due to the rapid growth and development of advanced driving assistance system. This paper uses the popularity of dash cam, developing the forward collision warning system based on monocular camera, and let the detection range in 50 meters to achieve real-time, stability and robustness as the goal.
The system algorithm is divided into four blocks, Cascade classifier training, vehicle detection, vehicle tracking and collision judgment. We use the Harr-features with adaboost algorithm and Active-learning Framework to train a good Cascade classifier. Vehicle detection is two-steps framework, in the first step, cascade classifier using multi-scale search window to find vehicle candidates, in the second step, further screening to reduce the false detection, and if we confirm as vehicle objects, get better vehicle width information, so that the final collision judgement has a better performance. Then recognizing vehicles in sequential frames, tracking them. Last, we find the most priority vehicle in collision region, calculating the time to contact with vehicle widths and sampling time, if less than two seconds, the driver will be warned.
We analysis consume time of each software algorithm, first-step vehicle detection is the most time-consuming part of the entire system, also the most front-end, so we select this part to design hardware architecture, to speed up the operation, the hardware architecture design including partition and scaler block, integral image generation block, line buffers block, search window registers block and parallel cascade classifier block. Partition and scaler block can greatly reduce the use of hardware resources. The use of two search window registers directly reduces a half of the overall processing time, and can also let the line buffers data have a higher reusability. Parallel cascade classifier block directly parallels the stages of cascade classifier, quickly determine whether the vehicle, to achieve our goal 60 FPS.

第一章 引言 1
一、車用電子 1
二、車用防撞系統 2
三、機器學習 3
四、論文組織 4
第二章 基於單鏡頭車輛偵測相關理論與文獻 5
一、單鏡頭車輛偵測 5
(一) 基於外觀方法 5
1. 特徵 5
2. 分類 7
(二) 基於行為方法 7
二、單鏡頭車輛偵測相關文獻介紹 8
(一) On-road Vehicle Detection: A Review [18] 8
(二) Active Learning for On-road Vehicle Detection: A Comparative Study [13] 9
(三) Symmetry-based Monocular Vehicle Detection System [8] 10
第三章 應用於單鏡頭前方碰撞預警系統演算法以及模擬結果 12
一、前言 12
二、串聯式分類器訓練 14
(一) 自適應增強分類器(Ada-boost Classifier) 14
1. 哈爾特徵(Haar-like Features) 14
2. 積分影像(Integral Image) 15
3. 自適應增強分類器訓練 18
(二) 串聯式分類器(Cascade of Classifiers) 20
(三) Active-Learning Framework 24
三、輸入畫面前處理 30
四、車輛偵測 31
(一) 多尺度視窗搜尋(Multi-Scale Search Window) 32
(二) 矩形邊界產生 36
五、車輛追蹤 40
(一) 歐幾里得距離計算 41
(二) 車輛追蹤的判定 42
六、碰撞判斷 43
(一) 計算碰撞興趣區塊中最優先的車輛 44
(二) 碰撞時間 44
七、模擬結果 47
(一) 量化數據定義 47
(二) 車輛偵測量化數據 48
(三) 兩階段車輛偵測數據分析 55
第四章 硬體架構設計與實作 56
一、前言 56
二、硬體規格 56
三、硬體架構設計 58
四、各單元之硬體架構設計 59
(一) 切割與縮放 59
(二) 積分影像產生 61
(三) Line Buffers 62
(四) 搜尋視窗暫存器 64
(五) 平行化之串聯式分類器 65
五、實作結果 67
(一) 數位IC之設計流程 67
(二) 晶片規格 69
(三) SYNTHESIS 71
(四) LAYOUT 72
第五章 結論 74
文獻參考 75

[1]李建汶, “視差圖產生之演算法研究及其電路架構設計” 國立中興大學電機工程學系碩士論文,中華民國103年1月。
[2]黃志銘, “移動物體與危險偵測之演算法研究及其電路架構設計與實現” 國立中興大學電機工程學系碩士論文,中華民國103年1月。
[3]楊俊成, “應用於行車紀錄器於2D to 3D轉換之深度知覺與測距演算法及消失線硬體架構設計” 國立中興大學電機工程學系碩士論文,中華民國101年6月。
[4]E. Dagan, O. Mano, G. P. Stein and A. Shashua, "Forward collision warning with a single camera," IEEE Intelligent Vehicles Symposium, 2004, 2004, pp. 37-42.
[5]G. P. Stein, O. Mano and A. Shashua, "Vision-based ACC with a single camera: bounds on range and range rate accuracy," IEEE IV2003 Intelligent Vehicles Symposium. Proceedings (Cat. No.03TH8683), 2003, pp. 120-125.
[6]Raphael, Eric, et al. "Development of a camera-based forward collision alert system." SAE International Journal of Passenger Cars-Mechanical Systems 4.2011-01-0579 (2011): 467-478.
[7]N. Srinivasa, "Vision-based vehicle detection and tracking method for forward collision warning in automobiles," Intelligent Vehicle Symposium, 2002. IEEE, 2002, pp. 626-631 vol.2.
[8]Teoh, Soo Siang, and Thomas Bräunl. "Symmetry-based monocular vehicle detection system." Machine Vision and Applications (2012): 1-12.
[9]X. Wang, L. Xu, H. Sun, J. Xin and N. Zheng, "On-Road Vehicle Detection and Tracking Using MMW Radar and Monovision Fusion," in IEEE Transactions on Intelligent Transportation Systems, vol. 17, no. 7, pp. 2075-2084, July 2016.
[10]A. Broggi, P. Cerri and P. C. Antonello, "Multi-resolution vehicle detection using artificial vision," IEEE Intelligent Vehicles Symposium, 2004, 2004, pp. 310-314.
[11]M. Cheon, W. Lee, C. Yoon and M. Park, "Vision-Based Vehicle Detection System With Consideration of the Detecting Location," in IEEE Transactions on Intelligent Transportation Systems, vol. 13, no. 3, pp. 1243-1252, Sept. 2012.
[12]S. Sivaraman and M. M. Trivedi, "Active learning based robust monocular vehicle detection for on-road safety systems," 2009 IEEE Intelligent Vehicles Symposium, Xi''an, 2009, pp. 399-404.
[13]Sivaraman, Sayanan, and Mohan M. Trivedi. "Active learning for on-road vehicle detection: A comparative study." Machine vision and applications (2014): 1-13.
[14]S. Sivaraman and M. M. Trivedi, "A General Active-Learning Framework for On-Road Vehicle Recognition and Tracking," in IEEE Transactions on Intelligent Transportation Systems, vol. 11, no. 2, pp. 267-276, June 2010.
[15]Viola, Paul, and Michael J. Jones. "Robust real-time face detection." International journal of computer vision 57.2 (2004): 137-154.
[16]P. Viola and M. Jones, "Rapid object detection using a boosted cascade of simple features," Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. CVPR 2001, 2001, pp. I-511-I-518 vol.1.
[17]Lienhart, Rainer, and Jochen Maydt. "An extended set of haar-like features for rapid object detection." Image Processing. 2002. Proceedings. 2002 International Conference on. Vol. 1. IEEE, 2002.
[18]Zehang Sun, G. Bebis and R. Miller, "On-road vehicle detection: a review," in IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 28, no. 5, pp. 694-711, May 2006.
[19]Zehang Sun, G. Bebis and R. Miller, "Monocular precrash vehicle detection: features and classifiers," in IEEE Transactions on Image Processing, vol. 15, no. 7, pp. 2019-2034, July 2006.
[20]J. Cui, F. Liu, Z. Li and Z. Jia, "Vehicle localisation using a single camera," 2010 IEEE Intelligent Vehicles Symposium, San Diego, CA, 2010, pp. 871-876.
[21]S. Sivaraman and M. M. Trivedi, "Looking at Vehicles on the Road: A Survey of Vision-Based Vehicle Detection, Tracking, and Behavior Analysis," in IEEE Transactions on Intelligent Transportation Systems, vol. 14, no. 4, pp. 1773-1795, Dec. 2013.
[22]S. Sivaraman and M. M. Trivedi, "A review of recent developments in vision-based vehicle detection," 2013 IEEE Intelligent Vehicles Symposium (IV), Gold Coast, QLD, 2013, pp. 310-315.
[23]M. Hiromoto, H. Sugano and R. Miyamoto, "Partially Parallel Architecture for AdaBoost-Based Detection With Haar-Like Features," in IEEE Transactions on Circuits and Systems for Video Technology, vol. 19, no. 1, pp. 41-52, Jan. 2009.
[24]Brousseau, Braiden, and Jonathan Rose. "An energy-efficient, fast FPGA hardware architecture for OpenCV-compatible object detection." Field-Programmable Technology (FPT), 2012 International Conference on. IEEE, 2012.
[25]H. C. Lai, M. Savvides and T. Chen, "Proposed FPGA Hardware Architecture for High Frame Rate (≫100 fps) Face Detection Using Feature Cascade Classifiers," 2007 First IEEE International Conference on Biometrics: Theory, Applications, and Systems, Crystal City, VA, 2007, pp. 1-6.
[26]C. Kyrkou and T. Theocharides, "A Flexible Parallel Hardware Architecture for AdaBoost-Based Real-Time Object Detection," in IEEE Transactions on Very Large Scale Integration (VLSI) Systems, vol. 19, no. 6, pp. 1034-1047, June 2011.
[27]R. C. Luo and H. H. Liu, "Design and implementation of efficient hardware solution based sub-window architecture of Haar classifiers for real-time detection of face biometrics," 2010 IEEE International Conference on Mechatronics and Automation, Xi''an, 2010, pp. 1563-1568.
[28]J. S. Lee, H. C. Chang and J. I. Guo, "An Adaboost-based two-level moving object detection architecture with dynamic ROI allocation," 2014 IEEE International Conference on Consumer Electronics - Taiwan, Taipei, 2014, pp. 103-104.
[29]M. Kimura, J. Matai, M. Jacobsen and R. Kastner, "A low-power Adaboost-based object detection processor using Haar-like features," 2013 IEEE Third International Conference on Consumer Electronics ¿¿ Berlin (ICCE-Berlin), Berlin, 2013, pp. 203-206.
[30]S. Park, S. Lee, S. Kim and K. Cho, "Design of AdaBoost classifier circuit using Haar-like features for automobile applications," 2011 International SoC Design Conference, Jeju, 2011, pp. 262-265.