臺灣博碩士論文加值系統

本研究的目的在於建構一個具有兩個迴轉自由度的光學追蹤系統，可以框選任意物體進行追蹤。為了有效的追蹤，並避免目標改變造成的追蹤失敗，使用的追蹤演算法具有學習的功能，能夠因應目標形態變化、被遮擋或消失等問題，並在追蹤失敗後，可以重新追回。
在研究中我們比較了三種追蹤演算法，分別為 CAMShift、Compressive Tracking 及 TLD (Tracking-Learning-Detection)。其中 CAMShift 的速度最快，但沒有學習的功能，無法在失去目標時重新追回。其次是 Compressive Tracking，利用壓縮降維提升計算速度，且擁有學習的功能，但由於測量尺度固定，影響目標大小改變時的追蹤效果，且不易追回被遮擋或暫時消失的目標。而 TLD 同時擁有追蹤、檢測與學習的能力，只要經過適當的學習，能夠擁有強健的追蹤能力，並能追回遺失的目標，但所需運算時間較長。
本研究以 TLD 為基礎，使用多執行緒平行化處理的方式提高其運算效能，追蹤畫面的更新速度可以達到每秒30張，並以實驗證實系統的可行性。

The goal of this thesis is to develop a dynamic optical tracking system mounted on a two-axis platform to track any selected object in video acquired by a camera. To enhance reliability, the developed tracking algorithm is equipped with learning capability to tackle the problems of morphological variation, disappearing and blocking of the target image.
Three well-known tracking algorithms are compared in this study: CAMShift, Compressive Tracking and TLD (Tracking Learning Detection). CAMShift has the benefit of simplicity and can be executed with higher sampling rate, but is lack of learning capability to adapt for variation in target image. Compressive Tracking algorithm employs dimensionality reduction to simplify computation, but is unable to cope with the problem of change in target image size due to fixed reference image size. TLD algorithm has the learning capability but requires heavy computation.
This study improves the execution efficiency of the system by implementing the TLD algorithm in multi-thread parallel processing. Experimental results show that the update rate can reach 30 frames per second and the system is able to actively track moving targets with learning capability.

指導教授推薦書
口試委員會審定書
授權書
致 謝 iv
中文摘要 v
英文摘要 vi
目錄 vii
圖表目錄 ix
第一章 緒論 - 1 -
1.1研究背景與動機 - 1 -
1.2文獻回顧 - 2 -
1.3研究目的 - 8 -
第二章 研究方法 - 9 -
2.1硬體架構 - 9 -
2.1.1工作站電腦 - 10 -
2.1.2影像擷取卡與攝影機 - 10 -
2.1.3運動控制卡與馬達驅動器 - 11 -
2.1.4二軸影像追蹤平台 - 13 -
2.1.5光遮斷器與I/O卡 - 15 -
2.1.6搖桿與觸控面板 - 16 -
2.2軟體架構 - 18 -
2.2.1 Microsoft Visual C++ 2008 - 18 -
2.2.2 OpenCV - 18 -
2.2.3 PISO-400 Windows DLL - 19 -
2.2.4 OpenMP - 19 -
2.3影像處理 - 20 -
2.3.1影像擷取 - 21 -
2.3.2追蹤演算法 - 21 -
第三章 實驗結果 - 38 -
3.1演算法比較 - 38 -
3.2演算法改良 - 60 -
3.3動態追蹤系統之實現 - 62 -
第四章 結論與未來方向 - 66 -
4.1結論 - 66 -
4.2未來研究方向 - 67 -
參考文獻 - 69 -

圖表目錄
圖1.1背景相減法[2] - 3 -
圖1.2 主動式輪廓偵測應用於交通監測[8,9] - 5 -
圖1.3 建立3D模型[11] - 6 -
圖2.1 ADlink RTV-24 影像擷取卡 - 10 -
圖2.2 光學變焦CCD攝影機 - 11 -
圖2.3 四軸步進/脈衝型伺服馬達控制卡 - 12 -
圖2.4 NIKI DD伺服馬達 - 12 -
圖2.5 三菱伺服馬達和J4驅動器 - 13 -
圖2.6 二軸影像追蹤機構設計圖 - 13 -
圖2.7 二軸影像追蹤機構細部分解 - 14 -
圖2.8 設計圖橫剖面 - 14 -
圖2.9 二軸影像追蹤機構實體圖 - 15 -
圖2.10 光遮斷器 - 16 -
圖2.11 觸控顯示器顯示影像 - 17 -
圖2.12 系統架構圖 - 20 -
圖2.13 直方圖顏色統計 - 23 -
圖2.14 顏色機率分布圖 - 23 -
圖2.15 Mean-Shift流程圖 - 24 -
圖2.16 CAMShift的流程 - 25 -
圖2.17 壓縮 - 28 -
圖2.18 更新分類器 - 30 -
圖2.19 Compressive Tracking流程 - 30 -
圖2.20 TLD演算法的組成 - 32 -
圖2.21 TLD的流程圖 - 32 -
圖2.22 TLD檢測器流程圖 - 35 -
圖3.1 CAMShift 框選目標與追蹤實現 - 39 -
圖3.2 CAMShift目標追蹤 - 40 -
圖3.3 與背景類似的目標追蹤 - 42 -
圖3.4 人臉目標發生遮擋交錯 - 43 -
圖3.5 Compressive Tracking 框選目標 - 44 -
圖3.6 Compressive Tracking目標追蹤實現 - 45 -
圖3.7 Compressive Tracking與背景類似的目標追蹤 - 48 -
圖3.8 Compressive Tracking人臉目標發生遮擋交錯 - 49 -
圖3.9 Compressive Tracking試圖重新追回 - 51 -
圖3.10 Compressive Tracking增加搜索範圍 - 52 -
圖3.11 TLD框選目標與追蹤實現 - 55 -
圖3.13 TLD人臉目標發生遮擋交錯 - 58 -
表3.1 演算法比較圖 - 60 -
圖3.14平行化 - 61 -
圖3.15 整合二旋轉軸追蹤平台和TLD演算法 - 63 -
圖3.16 測試目標側轉學習與檢測的過程 - 64 -
圖3.17 測試追蹤小目標 - 65 -
圖3.18 測試追蹤任意目標(箱子) - 65 -

[1] H. Weiming, T. Tieniu, W. Liang, and S. Maybank, "A survey on visual surveillance of object motion and behaviors," IEEE Transactions on Systems, Man, and Cybernetics, Part C: Applications and Reviews, vol. 34, no. 3, pp. 334-352, 2004.
[2] S. J. McKenna, S. Jabri, Z. Duric, A. Rosenfeld, and H. Wechsler, "Tracking groups of people," Computer Vision and Image Understanding, vol. 80, no. 1, pp. 42-56, Oct 2000.
[3] 鄒華龍，〈即時的人物偵測與追蹤〉，大同大學電機工程研究所，碩士論文，民國91年。
[4] A. Giachetti, "Matching techniques to compute image motion," Image and Vision Computing, vol. 18, no. 3, pp. 247-260, Feb. 2000.
[5] 蔡孟儒，〈應用即時視覺伺服系統實現平面輪廓追循〉，國立東華大學電機工程研究所，碩士論文，民國90年。
[6] 郭騰凱，〈在複雜背景下以強健視覺伺服為基礎可自動調變倍率之頭部追蹤系統〉，台灣大學電機工程學研究所，碩士論文，民國91年。
[7] T. K. Kuo , L. C. Fu, Jean, J.-H., P. Y. Chen, and Y. M. Chan, "Zoom-based head tracker in complex environment," in Proceedings of the 2002 International Conference on Control Applications, CA 2002, vol. 2, pp. 725-730, 2002.
[8] J. C. Tai, S. T. Tseng, C. P. Lin, and K. T. Song, "Real-time image tracking for automatic traffic monitoring and enforcement applications," Image and Vision Computing, vol. 22, Issue 6, pp. 485-501, June 2004.
[9] S. T. Tseng and K. T. Song, "Real-time image tracking for traffic monitoring," in Proceedings of the IEEE 5th International Conference on Intelligent Transportation Systems, 2002, pp. 1-6, 2002
[10] 王健權，〈以數位影像處理搭配動態攝影機做即時人臉追蹤之研究〉，國立成功大學電機工程學系，碩士論文，民國91年。
[11] Q. Delamarre and O. Faugeras, "3D articulated models and multiview tracking with physical forces," Computer Vision and Image Understanding, vol. 81, Issue 3, pp. 328-357, Mar. 2001.
[12] N. Paragios and R. Deriche, "Geodesic active contours and level sets for the detection and tracking of moving objects," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 22, no. 3, pp. 266-280, Mar. 2000.
[13] Z. Kalal, K. Mikolajczyk, and J. Matas, "Tracking-Learning-Detection," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 34, no. 1, pp. 1409-1422, Jul. 2012.
[14] G. Bradski, "The OpenCV library," Dr Dobbs Journal, vol. 25, pp. 120, Nov. 2000.
[15] B. Kwolek, "CamShift-based tracking in joint color-spatial spaces," in Proceedings of Computer Analysis of Images and Patterns, vol. 3691, pp. 693-700, Sep. 2005.
[16] Y. Cheng, "Mean shift, mode seeking, and clustering," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 17, no. 8, pp. 790-799, 1995.
[17] D. L. Donoho, "Compressed sensing," IEEE Transactions on Information Theory, vol. 52, Issue 3, pp. 1289-1306, Apr. 2006.
[18] 洪誥廷，〈應用機械視覺之動態追蹤系統〉，長庚大學機械工程研究所，碩士論文，民國93年。
[19] D. Chakraborty and D. Patra, "Real time object tracking based on segmentation and Kernel based method," in Proceedings of 2010 International Conference on Industrial and Information Systems (ICIIS), pp. 426-429, Aug. 2010.
[20] K. Zhang, L. Zhang , and M. H. Yang, "Real-Time Compressive Tracking," European Conference on Computer Vision (ECCV), Part III, pp. 866–879, 2012.