嵌入式雙核心平台實現即時前景偵測

以作者查詢圖書館館藏

、以作者查詢臺灣博碩士

、以作者查詢全國書目

、勘誤回報

、線上人數：42

、訪客IP：3.147.82.252

姓名

陳俊霖(Jyun-lin Chen) 查詢紙本館藏

畢業系所

資訊工程學系

論文名稱

嵌入式雙核心平台實現即時前景偵測
(Implement a Real-time Foreground Detection on a Dual-Core Embedded Platform)

相關論文

★ 使用視位與語音生物特徵作即時線上身分辨識	★ 以影像為基礎之SMD包裝料帶對位系統
★ 手持式行動裝置內容偽變造偵測暨刪除內容資料復原的研究	★ 基於SIFT演算法進行車牌認證
★ 基於動態線性決策函數之區域圖樣特徵於人臉辨識應用	★ 基於GPU的SAR資料庫模擬器：SAR回波訊號與影像資料庫平行化架構 (PASSED)
★ 利用掌紋作個人身份之確認	★ 利用色彩統計與鏡頭運鏡方式作視訊索引
★ 利用欄位群聚特徵和四個方向相鄰樹作表格文件分類	★ 筆劃特徵用於離線中文字的辨認
★ 利用可調式區塊比對並結合多圖像資訊之影像運動向量估測	★ 彩色影像分析及其應用於色彩量化影像搜尋及人臉偵測
★ 中英文名片商標的擷取及辨識	★ 利用虛筆資訊特徵作中文簽名確認
★ 基於三角幾何學及顏色特徵作人臉偵測、人臉角度分類與人臉辨識	★ 一個以膚色為基礎之互補人臉偵測策略

檔案

[Endnote RIS 格式]

[Bibtex 格式]

[相關文章]

[文章引用]

[完整記錄]

[館藏目錄]

[檢視]

[下載]

本電子論文使用權限為同意立即開放。
已達開放權限電子全文僅授權使用者為學術研究之目的，進行個人非營利性質之檢索、閱讀、列印。
請遵守中華民國著作權法之相關規定，切勿任意重製、散佈、改作、轉貼、播送，以免觸法。

摘要(中)

監控系統廣泛的應用在日常生活中，不過目前所使用的監控系統
大都只有錄影功能，無法立即針對異常事件發出警訊，所以有人提出
智慧型監控系統的概念，利用電腦視覺的方式，在不需要人為操作的
方式下，讓監控系統能自動分析攝影機擷取的影像，使其具有偵測、
追蹤與辨識等功能。
本論文提出將前景偵測實現在嵌入式雙核心平台上，可以即時且
有效的偵測出前景物。前景偵測演算法的部份，我們是採用計算量低，記憶體使用量少的CodeBook 演算法；嵌入式平台實現的部分包含：軟體上利用Codec Engine 架構的方法，搭配在硬體上利用共享記憶體(share memory)的配置，來實現ARM 與DSP 彼此間資料的傳遞；在定點式DSP 上將程式定點化，來提高DSP 在處理資料上的效率；利用嵌入式平台雙核心特性設計多執行緒，來達到平行處理的效果。實驗結果可以顯示，本論文可以即時且有效的偵測出前景物，在
執行速度上每秒可以處理28 至30 張幀(frame)。

摘要(英)

Surveillance systems are widely applied in our daily life. However, the main function of current surveillance systems only focuses on video data recording. Besides, human has to pay tremendous attention in monitoring the video data and issuing alarm when abnormal events happen. To save the human resource, an automatic and intelligent surveillance system is developed in this thesis to detect, track, recognize and analyze moving objects.
The main goal of this thesis is to implement the task of foreground detection on a dual-core embedded platform. A codebook-based algorithm is developed to perform the moving object detection task in resource lacking workspace. Hence, lower computational power and less memory space are needed in performing the detection task. The main task in the dual-core system is to establish a reliable communication between ARM and DSP. A software-based core engine framework is adopted to allocate the share memory and support the hardware. Due to the non-supporting of floating-point arithmetic hardware in DSP, the fixed-point operation is performed to make DSP work efficiently. The resources are well allocated to execute the parallel process on the dual-core platform. The system is installed in two places to evaluate the performance. Experimental results demonstrate that the implemented system can accurately and efficiently detect the moving objects in real time.

關鍵字(中)

★ 異質雙核心處理器
★ 前景偵測
★ 嵌入式系統

關鍵字(英)

★ embedded system
★ foreground detection
★ heterogeneous dual-core processors

論文目次

Abstract ........................................................................................................I
摘要..................................................................................................... II
目錄................................................................................................... III
附圖目錄....................................................................................................VI
表格目錄....................................................................................................IX
第 1 章緒論................................................................................................ 1
1.1. 研究動機.......................................................................................... 1
1.2. 研究目的.......................................................................................... 2
1.3. 相關研究.......................................................................................... 3
1.4. 系統流程.......................................................................................... 4
1.5. 論文架構.......................................................................................... 6
第 2 章文獻探討........................................................................................ 7
2.1. 色彩空間.......................................................................................... 7
2.1.1. YCbCr色彩空間........................................................................................... 7
2.1.2. YCbCr轉換RGB........................................................................................... 8
2.2. 移動物體偵測.................................................................................. 9
2.2.1. CodeBook前景偵測法............................................................................... 10
2.2.2. 色彩失真與亮度失真............................................................................... 11
2.2.3. 背景模組建立........................................................................................... 15
2.2.4. 前景偵測................................................................................................... 17
2.3. 形態學影像處理............................................................................ 19
2.4. 連通元件標記................................................................................ 22
第 3 章軟硬體開發環境與流程.............................................................. 24
3.1. 硬體環境........................................................................................ 24
3.1.1. 雙核心嵌入式平台DM6446 EVM........................................................... 25
3.1.2. ARM ........................................................................................................... 25
3.1.3. DSP ............................................................................................................. 26
3.1.4. 彩色LCD螢幕........................................................................................... 28
3.1.5. 紅外線攝影機........................................................................................... 28
3.1.6. JTAG與ICE ................................................................................................ 29
3.2. 軟體環境........................................................................................ 30
3.2.1. Linux作業系統與交叉編譯器................................................................... 30
3.2.2. Code Composer Studio (CCS).................................................................... 30
3.2.3. XDC-Tool ................................................................................................... 30
3.3. 軟體開發流程................................................................................ 31
第 4 章雙核心嵌入式系統實現.............................................................. 32
4.1. ARM與DSP之間的溝通機制........................................................ 32
4.1.1. Code Engine架構........................................................................................ 33
4.1.2. 共享記憶體配置....................................................................................... 34
4.2. 定點化............................................................................................ 37
4.2.1. 定點化步驟............................................................................................... 37
4.2.2. 定點化的困難........................................................................................... 38
4.3. 平行化設計.................................................................................... 40
4.3.1. 擷取影像................................................................................................... 41
4.3.2. 演算法處理............................................................................................... 41
4.3.3. 顯示影像................................................................................................... 42
4.3.4. 終止控制................................................................................................... 42
4.4. 緩衝區記憶體配置........................................................................ 43
第 5 章效能測試與實驗結果.................................................................. 45
5.1. 執行速度分析................................................................................ 45
5.2. 背景建立........................................................................................ 46
5.3. 前景偵測........................................................................................ 47
5.3.1. 實驗室A .................................................................................................... 48
5.3.2. 實驗室B..................................................................................................... 49
5.3.3. 走廊A ........................................................................................................ 50
5.3.4. 走廊B......................................................................................................... 51
5.4. 效能比較與實驗結論.................................................................... 52
5.4.1. 效能比較................................................................................................... 52
5.4.2. 實驗結論................................................................................................... 53
第6 章結論與未來展望.......................................................................... 56
6.1. 結論................................................................................................ 56
6.2. 未來展望........................................................................................ 57
参考文獻.................................................................................................... 58

參考文獻

[1] Kyungnam Kim, Thanarat H. Chalidabhongse, David Harwood,
LarryDavis, “Real-time foreground–background segmentation using codebook model” Real-Time Imaging, Vol. 11, pp. 172-185, June 2005.
[2] Stauffer C, Grimson WEL. “Adaptive background mixture models for real-time tracking.” IEEE International Conference on Computer Vision and Pattern Recognition. Vol. 2, pp.246-252, 1999.
[3] Du-Ming Tsai and Shia-Chih Lai. “Independent Component
Analysis-Based Background Subtraction for Indoor Surveillance.” IEEE Transaction on Image Processing. Vol.18, No.1, January 2009 [4] Gonzalez Woods, 繆紹鋼編譯, 數位影像處理, 台灣培生教育出版 2003
[5] TI, TMS320DM644x DMSoC ARM Subsystem, Literature Number
SPRUE14A, March 2007
[6] TI, TMS320C64x DSP CPU and Instruction Set, Literature Number SPRU732H, October 2008
[7] 盧怡仁, 蔣偉和, 單晶片於數位信號處理的應用, 文魁出版, 2007
[8] TI, Code Composer Studio Development Tool v3.3 Getting Started Guide, Literature Number SPRUE509H, October 2006
[9] TI, ARM Compiler Tips and Code Size Optimization Using
DSP/BIOS Link, Literature Number SPRUAAC3, November 2005
[10] TI, XDC Users Guide
[11] TI, Creating a TMS320DM6446 Audio Encode Example Using XDC Tools, Literature Number SPRAAI6A, February 2008
[12] TI, XDC Getting Started Guide, March 26 2008
[13] TI, Codec Engine Application Developer’s Guide, Literature Number SPRUE67D, September 2007
[14] TI, Codec Engine Algorithm Creator User’s Guide, Literature Number SPRUED6C, September 2007
[15] TI, Build a Small Embedded Linux Kernel Example, Literature Number SPRUAH2A, May 2008
[16] TMS320C6000 DSPBIOS User's Guide, Literature Number
SPRU303B, May 2000 59
[17] TI, DVSDK Getting Started Guide, Literature Number SPRUEG8, August 2006
[18] Jonathan Corbet, Alessandro Rubini, 林長毅編譯, Linux驅動程式, 歐萊禮出版, 2001
[19] Karim Yaghmour, 蔣大偉編譯, 建構嵌入式Linux系統, 歐萊禮出版, 2004
[20] 陳宏宇，DSP程式設計，松崗出版, 2004
[21] Video for Linux Two API Specification Draft 0.12，取自
http://v4l.videotechnology.com
[22] 鳥哥, 鳥哥的Linux私房菜, 上奇出版, 2002
[23] Banto ARM Linux，PIPE FIFO API。取自
http://bantolinux.blogspot.com/2007/05/pipe-fifo-api-v1.html

指導教授

范國清(Kuo-Chin Fan)

審核日期

2010-7-22

推文