(3.210.184.142) 您好!臺灣時間:2021/05/12 04:13
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果

詳目顯示:::

我願授權國圖
: 
twitterline
研究生:官振安
研究生(外文):Zhen-AnGuan
論文名稱:基於現場可程式化邏輯閘陣列之行人偵測與追蹤系統: 低記憶體成本方案
論文名稱(外文):A Real-Time FPGA-Based Pedestrian Detection and Tracking System: A Low Memory Cost Approach
指導教授:陳進興陳進興引用關係
指導教授(外文):Chin-Hsing Chen
學位類別:碩士
校院名稱:國立成功大學
系所名稱:電腦與通信工程研究所
學門:工程學門
學類:電資工程學類
論文種類:學術論文
論文出版年:2020
畢業學年度:108
語文別:英文
論文頁數:77
中文關鍵詞:現場可程式化邏輯閘陣列即時移動物體偵測移動物體辨識移動物體追蹤背景去除K-平均群集KBMOTKRBMOT
外文關鍵詞:FPGAreal-timemoving object detectionmoving object recognitionmoving object trackingbackground subtractionK-Means ClusteringKBMOTKRBMOT
相關次數:
  • 被引用被引用:0
  • 點閱點閱:43
  • 評分評分:系統版面圖檔系統版面圖檔系統版面圖檔系統版面圖檔系統版面圖檔
  • 下載下載:0
  • 收藏至我的研究室書目清單書目收藏:0
本文提出了一種基於現場可程式化邏輯閘陣列(FPGA)之即時移動物體偵測和追蹤系統,以行人為主要目標。從攝影機捕捉到的輸入視訊,經過如:彩色轉灰階、背景去除、形態學運算、K-方形氣球移動物體追蹤(KRBMOT)、邊框產生等一系列的處理之後,產生出能在視訊圖形陣列(VGA)螢幕上顯示的輸出視訊。整體的視訊處理能以超視訊圖形陣列(SVGA)的解析度,即 800×600 的解析度輸出,同時達到即時的要求,更準確地說,每秒60幀的影格率(FPS)。
本文提出的演算法KRBMOT是受了K-平均群集演算法(KMC)的啟發。KRBMOT是一個移動物體追蹤(MOT)演算法,藉由加入額外的參數到KMC再加上一些調整而得。KRBMOT特別適用於形狀接近長方形的物體,其中一個重要的應用就是行人追蹤。
如同KMC,KRBMOT只需要記憶體用來儲存很少的參數。這大大降低了本文提出的系統的記憶體需求。本文提出的系統只需要210 kbit的晶片內記憶體和一個頻寬為400 MB/s的晶片外同步動態隨機存取記憶體(SDRAM)。相較於其他方案,本文提出的系統建立在便宜得多的硬體架構上,同時維持良好的性能。
This thesis proposed a real-time moving object detection and tracking system, which mainly focuses on the pedestrians, based on a field programmable gate array (FPGA). After being applied a series of processing, including color to greyscale transformation, background subtraction, morphological denoising, the K-Rectangle-Balloons Moving Object Tracking (KRBMOT), and bounding boxes generation, the input color video captured by the camera is transformed to the output color video, which is shown on the video graphics array (VGA) screen. The proposed video processing system is real-time, specifically, its frame rate is 60 frames per second (FPS), each of resolution 800×600.
The proposed algorithm, KRBMOT, is inspired by the K-Means Clustering (KMC) algorithm. KRBMOT is a moving object tracking (MOT) algorithm derived by adding additional parameters to KMC with a few adjustments. KRBMOT is suitable for objects with shape closed to a rectangle. An important application of KRBMOT is pedestrian tracking. Like KMC, KRBMOT only requires memories to store very few parameters. This greatly reduces the memories required by the proposed system. The total memories which the proposed system used are 210 kbit on-chip memories and an off-chip synchronous dynamic random-access memory (SDRAM) with 400 MB/s bandwidth. Comparing to other approaches, the proposed system is much cheaper while with good performance.
摘 要 I
Abstract III
誌 謝 V
Acknowledgment VI
Contents VII
List of Tables X
List of Figures XI
Chapter 1 Introduction 1
1.1 Background 1
1.1.1 Field Programmable Gate Array (FPGA) 1
1.1.2 Hardware Architecture Design (HAD) 1
1.1.3 Moving Object Detection (MOD) and Moving Object Tracking (MOT) 2
1.2 Motivation 4
1.2.1 Choice of Background Subtraction Algorithm 4
1.2.2 The Proposed MOT Algorithm 5
1.3 Thesis Contribution 6
1.4 Thesis Outline 6
Chapter 2 Models and Algorithms 7
2.1 Overview 7
2.2 Tensor Formulation for Videos 9
2.3 Color to Greyscale Transformation 13
2.4 Background Subtraction (BS) 14
2.5 Morphological Denoising 15
Chapter 3 The Proposed Moving Object Tracking Algorithm 18
3.1 Overview 18
3.2 The K-Means Clustering Algorithm (KMC) 19
3.3 Problems in Extending KMC to MOT 22
3.4 The K-Balloons MOT Algorithm (KBMOT) 24
3.5 Object Rediscovering and Merging 28
3.6 The K-Rectangle-Balloons MOT Algorithm (KRBMOT) 33
3.7 Tensorization for KRBMOT 36
3.8 Bounding Boxes Generation 43
Chapter 4 Hardware Architecture Design of the Proposed System 49
4.1 Devices and Environments 49
4.1.1 The DE2-115 Board 49
4.1.2 The TRDB-D5M Camera Kit 49
4.1.3 Programming Interface 49
4.1.4 The Cyclone IV E FPGA 51
4.1.5 SDRAM 51
4.1.6 VGA DAC 51
4.1.7 RS232 Interface 51
4.2 The Top-Level Module and Data Transmission 53
4.2.1 The Top-Level Module 53
4.2.2 Tensor Transmission 53
4.2.3 Notations Explanation 55
4.3 The Color to Greyscale Transformation Module 57
4.4 The Parameter-Update Module 59
4.5 The Background Subtraction Module 63
4.6 The Morphological Denoising Module 66
4.7 The KRBMOT Module 67
Chapter 5 Experimental Results 69
Chapter 6 Conclusion 74
References 75
[1]D. G. Bailey, Design for Embedded Image Processing on FPGAs, John Wiley & Sons (Asia), Singapore, SG, pp. 21-25, 234-248, 352-359, 2011.
[2]F. Chollet, Deep Learning with Python, Manning Publications, Shelter Island, NY, US, pp. 31-37, 2018.
[3]S. Friedberg, A. Insel, and L. Spence, Linear Algebra, 4 ed., Pearson Education, London, England, UK, pp. 439-443, 2014.
[4]M. Genovese, E. Napoli, D. De Caro, N. Petra, and A. G. M. Strollo, FPGA Implementation of Gaussian Mixture Model Algorithm for 47 fps Segmentation of 1080p Video, Journal of Electrical and Computer Engineering, vol. 2013, Art no. 129589, pp. 1-8, 2013.
[5]R. C. Gonzalez and R. E. Woods, Digital Image Processing, 4 ed., Pearson Education, London, England, UK, pp. 636-648,770-772, 2018.
[6]J. Heikkilä and O. Silven, A real-time system for monitoring of cyclists and pedestrians, in Proceedings Second IEEE Workshop on Visual Surveillance (VS'99) (Cat. No.98-89223), pp. 74-81, 1999.
[7]J. Hoshen and R. Kopelman, Percolation and cluster distribution. I. Cluster multiple labeling technique and critical concentration algorithm, Phys. Rev. B, vol. 14, pp. 3438-3445, 1976.
[8]Integrated Silicon Solution Inc, IS42S86400B, IS42S16320B, IS45S16320B, Integrated Silicon Solution Inc, 2009.
[9]J. S. S. Kutty, F. Boussaid, and A. Amira, A high speed configurable FPGA architecture for k-mean clustering, in 2013 IEEE International Symposium on Circuits and Systems (ISCAS), pp. 1801-1804, 2013.
[10]T. Lin, P. Goyal, R. Girshick, K. He, and P. Dollár, Focal Loss for Dense Object Detection, in 2017 IEEE International Conference on Computer Vision (ICCV), pp. 2999-3007, 2017.
[11]J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, You Only Look Once: Unified, Real-Time Object Detection, in 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 779-788, 2016.
[12]K. H. Rosen, Discrete Mathematics and Its Applications, McGraw-Hill,, New York, NY, US, pp. A3-1-A3-6, 2007.
[13]S. Shaikh, K. Saeed, and N. Chaki, Moving Object Detection Using Background Subtraction (SpringerBriefs in Computer Science), Springer, Heidelberg, DE, pp. 5-14, 2014.
[14]S. Shalev-Shwartz and S. Ben-David, Understanding Machine Learning, Cambridge University Press, Cambridge, England, UK, pp. 268-271, 2014.
[15]C. Stauffer and W. E. L. Grimson, Adaptive background mixture models for real-time tracking, in Proceedings. 1999 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (Cat. No PR00149), vol. 2, pp. 246-252, 1999.
[16]Terasic Technology Inc., DE2-115 User Manual, Terasic Technology Inc., 2012.
[17]Terasic Technology Inc., TRDB-D5M Hardware Specification, Terasic Technology Inc., 2009.
[18]Terasic Technology Inc., TRDB-D5M User Manual, Terasic Technology Inc., 2017.
[19]S. Zhang, Real Time Image Processing on FPGAs, PhD thesis, Department of Electrical Engineering and Electronics, University of Liverpool, Liverpool, England, UK, pp. 67-71, 128-144, 2018.
[20]Zywyn Corparation, ZT3220E, ZT3221E, ZT3222E, ZT3223E, ZT3232E, ZT1385E, Zywyn Corparation, 2005.
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top
系統版面圖檔 系統版面圖檔