簡易檢索 / 詳目顯示
| 研究生： |
羅善寬 Shan-Kuan Lo |
|---|---|
| 論文名稱： |
高斯混合模型之移動物體偵測實現於FPGA Gaussian Mixture Model for Moving Object Detection Implemented on FPGA |
| 指導教授： |
王乃堅
Nai-Jian Wang |
| 口試委員: |
呂學坤
鍾順平 姚嘉瑜 郭景明 王乃堅 |
| 學位類別： |
碩士 Master |
| 系所名稱： |
電資學院 - 電機工程系 Department of Electrical Engineering |
| 論文出版年： | 2024 |
| 畢業學年度： | 112 |
| 語文別： | 中文 |
| 論文頁數： | 63 |
| 中文關鍵詞： | 移動物體偵測 、背景減除法 、高斯混合模型 、FPGA 、即時 |
| 外文關鍵詞： | Moving Object Detection, Background Subtraction, Gaussian Mixture Model, FPGA, Real-time |
| 相關次數： | 點閱：159 下載：4 |
| 分享至: |
| 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 |
影像監控中，移動物體偵測與背景濾除極為重要。消除背景並提取前景以利後續的分析，例如行為分析、物件追蹤等。隨著自駕車技術興起，移動物體偵測也漸漸受到關注。
傳統移動物體偵測的方法分為三類，分別為幀差法(Frame Differencing)、背景濾除法(Background Subtraction)和光流法(Optical Flow)。幀差法：背景不變的情況下，兩幀影像有不同之處則為前景，因此將兩幀影像進行差分運算來得出移動物體；背景濾除法：使用估計、統計或建模等方式，將前景與背景區分開來。光流法：如果圖像中没有移動物體，光流場連續均勻，反之，移動物體會對光流場造成改變，觀察相鄰幀的光流場變化來得出移動物體。
本篇論文基於背景濾除法，以高斯混合模型(GMM)算法為基礎，並對其進行改良，最後再將改良的算法實現在FPGA(Field Programmable Gate Array)上。改良的算法減少記憶體、計算上和模型參數的更新上的成本，且能做出與GMM相近的結果。由於此演算法製作簡易，因此能很好的實現於FPGA。
當硬體接收到影像序列的輸入，首先將影像進行預處理，並從預處理後的影像進行高斯分類，並對高斯中的參數逕行迭代更新，以此更新背景模型，最後判斷是否為前景背景，已偵測移動物體。
實驗結果顯示此方法可成功從攝影鏡頭輸入並將移動物體偵測後的影像顯示輸出至螢幕上，且最後的結果顯示本系統使用了1,991個邏輯元件和1,667,662 bits內部記憶體，功耗為371.97mW，且處理速度為71.44mu s延遲(NTSC Input)，達到即時輸出的效果。
關鍵詞: 移動物體偵測、背景減除法、高斯混合模型、FPGA、即時
In image surveillance, moving object detection and background subtraction are essential for subsequent analysis like behavior analysis and object tracking. With the rise of autonomous vehicles, this has gained increased attention.
Traditional methods for moving object detection include Frame Differencing, Background Subtraction, and Optical Flow. Frame Differencing detects changes between two frames to identify moving objects. Background Subtraction distinguishes foreground from background using statistical methods. Optical Flow identifies moving objects by observing changes in the optical flow field between adjacent frames.
This thesis improves upon the Background Subtraction method using the Gaussian Mixture Model (GMM) algorithm, implementing the improved algorithm on FPGA (Field Programmable Gate Array). The improvements reduce memory, computational, and parameter update costs, while maintaining similar accuracy to GMM. This simplified algorithm is well-suited for FPGA implementation.
The hardware preprocesses the input image sequence, performs Gaussian classification, updates Gaussian parameters iteratively, and determines the foreground or background to detect moving objects.
Experimental results demonstrate successful real-time moving object detection with a system using 1,991 logic elements, 1,667,662 bits of internal memory, 371.97 mW power consumption, and a processing speed of 71.44mu s latency (NTSC Input).
Keywords: Moving Object Detection, Background Subtraction, Gaussian Mixture Model, FPGA, Real-time
[1] S.S. Sengar, S. Mukhopadhyay, “A novel method for moving object detection based on block based frame differencing”, 2016 3rd International Conference on Recent Advances in Information Technology (RAIT), IEEE, 2016.
[2] B.P.L. Lo, S.A. Velastin, “Automatic congestion detection system for underground platforms”, Proceedings of 2001 International Symposium on Intelligent Multimedia, Video and Speech Processing. ISIMP 2001, May 2001.
[3] R. Cucchiara, C. Grana, M. Piccardi, “Detecting moving objects, ghosts, and shadows in video streams”, IEEE Transactions on Pattern Analysis and Machine Intelligence, Volume 25, Issue 10, October 2003.
[4] F.E. Baf, T. Bouwmans, B. Vachon, “Fuzzy integral for moving object detection”, 2008 IEEE International Conference on Fuzzy Systems, June 2008.
[5] F.E. Baf, T. Bouwmans, B. Vachon, “A fuzzy approach for background subtraction”, 15th IEEE International Conference on Image Processing, October 2008.
[6] A.Elgammal, D.Harwood, L.Davis, “Non-parametric model for background subtraction”, ECCV 2000, LNCS 1843, 2000, pp. 751–767.
[7] D.S. Lee, “Effective Gaussian mixture learning for video background subtraction”, IEEE Transactions on Pattern Analysis and Machine Intelligence, Volume 27, Issue 5, May 2005.
[8] A. Ilyas, M.Scuturici, S.Miguet, “Real time foreground-background segmentation using a modified codebook model”, 2009 Sixth IEEE International Conference on Advanced Video and Signal Based Surveillance, September 2009.
[9] Y. Benezeth, P.M. Jodoin, B. Emile, H. Laurent, C. Rosenberger, “Comparative study of background subtraction algorithms”, Journal of Electronic Imaging, July 2010.
[10] B.K.P. Horn, B.G. Schunck, “Determining optical flow”, Artificial Intelligence, 1981, pp. 185-204.
[11] D.R. Gilland, B.A. Mair, “Motion estimation in gated cardiac emission tomography by optical flow techniques”, IEEE Nuclear Science Symposium Conference Recor, 2006, pp. 2699-2702.
[12] Y. Wang, P.M. Jodoin, F. Porikli, J. Konrad, Y. Benezeth, and P. Ishwar, “CDnet 2014: An expanded change detection benchmark dataset”, 2014 CVPR IEEE Workshop on Change Detection (CDW-2014), 2014, pp. 387-394.
[13] G.Takhar, C.Prakash, N.Mittal, R.Kumar, “Comparative analysis of background subtraction techniques and applications”, 2016 International Conference on Recent Advances and Innovations in Engineering (ICRAIE), June 2017.
[14] S.Elhabian, K.M. El-Sayed, S.H. Ahmed, “Moving object detection in spatial domain using background removal techniques-state-of-art”, Recent Patents on Computer Science, January 2008, 1, pp. 32-54.
[15] S.Y Chiu, C.C Chiu, S.D Xu, “A Background Subtraction algorithm in complex environments based on Category Entropy analysis”, Advanced Internet of Things for Smart Infrastructure System, May 2018.
[16] Terasic DE-115 datasheet URL:http://www.terasic.com.tw/tw/
[17] Sony EVI-D70 技術手冊 URL:http://pro.sony/en_GR/products/ptz-network-cameras/evi-d70-d70p-pal-
