臺灣博碩士論文加值系統

本篇論文主要提出經由霍夫轉換(Hough Transform)之頭部偵測演算法後，得
到初步的頭部候選者，再經由整合形狀資訊、顏色資訊與前景資訊做進一步的頭
部候選者驗證。在本論文中，首先針對場景作定義，在攝影機固定的情況下，我
們假設當攝影機為一般監控環境下所拍攝的場景，在此情況人體頭部的區塊較不
容易有遮蔽的情況發生，所以我們以提取影像中每個人形的頭部區塊作為人群偵
測的依據。由於頭部的輪廓並非完整的圓形，而且會受到頭髮、帽子或是其他物
體的部分遮蔽等影響，使得霍夫轉換(Hough Transform)找出來的候選者錯誤率很
高，所以需要再經過後面第二階段的驗證，才能將錯誤的候選者排除。最後在實
驗的部分，會展示在不同情況下本篇方法的偵測結果。

This paper proposes a vision-based crowd segmentation method based on a single
camera. The main idea behind our work is to fuse multiple cues so that the major
challenges, such as occlusion and complex background facing in the crowd
segmentation area can be successfully overcome. Based on the assumption that the
human heads are visible in the image, we use the circle hough transform (CHT) to
detect all circular regions and all of them are considered as candidates of human heads.
Due to the effects from complex background or clothes, lots of those detected
candidates are false positives. Accordingly, we firstly apply template matching (TM) to
remove the false candidates at the lower half part of human body. Then, two proposed
cues called head foreground contrast (HFC) and block color relation (BCR) are
incorporated for further verification. Every candidate that passes through these tests is
considered as a human in an image. Finally, the geometric relationships as well as the
mask from background subtraction are used to perform the segmentation. In experiment,
three videos are used to validate the proposed approach. Our proposed method
effectively lowers the false positives with only sacrificing little detection rate.

中文摘要………………………………………………………………………………I
英文摘要………………………………………………………………………………II
致謝…………………………………………………………………………………III
目錄……………………………………………………………………………………IV
表目錄……………………………………………………………………………VII
圖目錄………………………………………………………………………………VIII
第一章 緒論…………………………………………………………………………1
1.1. 研究動機……………………………………………………………………1
1.2. 困難與挑戰……………………………………………………………………2
1.2.1 遮蔽問題………………………………………………………………2
1.2.2 複雜背景………………………………………………………………3
1.2.3 相機監控視角……………………………………………………………4
1.2.4 人體姿態高自由度……………………………………………………4
1.3. 相關研究……………………………………………………………………5
1.4. 系統概述………………………………………………………………………7
第二章 前景切割……………………………………………………………………11
2.1. 移動物體切割………………………………………………………………11
2.2. 前景切割……………………………………………………………………12
2.2.1. 初始化……………………………………………………………13
2.2.2. 背景相減…………………………………………………………13
2.2.3. 更新背景模型………………………………………………………14
第三章 基於霍式轉換之頭部偵測…………………………………………………15
3.1. 頭部偵測……………………………………………………………………15
V
3.2. 邊緣影像結合前景資訊……………………………………………………16
3.3. 霍式轉換……………………………………………………………………17
第四章 多重資訊整合驗證演算法…………………………………………………19
4.1.樣板比對……………………………………………………………………19
4.1.1. 距離轉換-(Distance Transform，DT)…………………………21
4.1.2. 樣板資料庫建立……………………………………………………23
4.1.3. 樣板比對……………………………………………………………23
4.2. 頭部前景對比-(Head Foreground Contrast，HFC)………………………26
4.2.1. 頭部前景對比估算…………………………………………………27
4.2.2. 無法處理情況…………………………………………………………27
4.3. 區塊顏色關聯性驗證-(Block Color Relation Verification，BCR)……28
4.3.1. 色彩資訊分析…………………………………………………………27
4.3.2. 色彩資訊判斷…………………………………………………………28
4.3.3. 無法處理情況…………………………………………………………29
4.4. 方法整合……………………………………………………………………31
4.5. 切割…………………………………………………………………………32
第五章 實驗結果與討論……………………………………………………………33
5.1. 實驗環境……………………………………………………………………33
5.2. 影片資料庫…………………………………………………………………34
5.3. 實驗設計……………………………………………………………………34
5.4. 實驗分數……………………………………………………………………35
5.5. 實驗結果……………………………………………………………………36
VI
5.6. 整合分析……………………………………………………………………40
5.7. 無法正常偵測情況…………………………………………………………42
第六章結論與未來展望…………………………………………………………43
參考文獻………………………………………………………………………………44
VII
表目錄
表5-1 實驗環境……………………………………………………………………33
表5-2 各階段執行時間(秒)………………………………………………………33
表5-3 影片資料庫…………………………………………………………………34
表5-4 Video1 偵測結果…………………………………………………………36
表5-5 Video2 偵測結果…………………………………………………………36
表5-6 Video3 偵測結果…………………………………………………………36
VIII
圖目錄
圖1-1 遮蔽問題……………………………………………………………………2
圖1-2 複雜背景示意圖……………………………………………………………3
圖1-3 不同的監控視角………………………………………………………………4
圖1-4 人體姿態變化………………………………………………………………5
圖1-5 因遮蔽問題而受損的邊緣資訊………………………………………………6
圖1-6 個視角示意圖………………………………………………………………7
圖1-7 切割示意圖…………………………………………………………………8
圖1-8 人群切割系統架構圖…………………………………………………………8
圖2-1 連續影像相減架構圖…………………………………………………………11
圖2-2 背景相減架構圖……………………………………………………………12
圖2-3 前景影像…………………………………………………………………14
圖3-1 不同觀點之頭部輪廓………………………………………………………15
圖3-2 Canny 邊緣影像………………………………………………………16
圖3-3 整合前景資訊之邊緣影像……………………………………………………17
圖3-4 參數定義……………………………………………………………………18
圖3-5 高低門檻值之霍式轉換偵測比較…………………………………………18
圖4-1 初步候選者…………………………………………………………………20
圖4-2 驗證階段架構圖…………………………………………………………20
圖4-3 樣板比對架構圖……………………………………………………………21
圖4-4 距離轉換範例圖……………………………………………………………22
IX
圖4-5 距離轉換……………………………………………………………………22
圖4-6 樣板資料庫…………………………………………………………………23
圖4-7 樣板比對範例………………………………………………………………24
圖4-8 樣板比對結果…………………………………………………………………25
圖4-9 前景區域驗證架構圖………………………………………………………26
圖4-10 頭部前景對比影像分析………………………………………………………26
圖4-11 無法處理情況示意圖………………………………………………………28
圖4-12 前景區域驗證結果…………………………………………………………28
圖4-13 區塊驗證架構圖……………………………………………………………28
圖4-14 相鄰區塊示意圖……………………………………………………………29
圖4-15 分數計算方式………………………………………………………………30
圖4-16 區塊驗證結果………………………………………………………………31
圖4-17 整合示意圖………………………………………………………………32
圖4-18 人體區域計算………………………………………………………………32
圖4-19 切割結果………………………………………………………………32
圖5-1 偵測結果示意圖……………………………………………………………34
圖5-2 嚴重頭部遮蔽示意圖……………………………………………………35
圖5-3 標記方法示意圖…… ………… … …………… ……… … …………35
圖5 - 4 V id e o1 偵測結果…… … …… …… … … …… … …… …… … … …37
圖5 - 5 V id e o1 偵測結果……… …… …… … … …… … …… …… … … …38
圖5 - 6 V id e o1 偵測結果…… … …… …… … … …… … …… …… … … …39
圖5-7 互補分析數據圖………………………………………………………41
X
圖5-8 無法正確偵測……………………………………………………………42

Real-Time Surveillance of People
and Their Activities ” , IEEE Transactions on Pattern Analysis and Machine
Intelligence, vol. 22, no. 8, pp. 809 - 830, 2002.
[2] Hedvig, Swedish Defence Res. Agency, “Detecting Human Motion with
Support Vector Machines”, International Conference on Pattern Recognition, pp.
188 - 191, 2004.
[3] Junghye Min, Rangachar Kasturi “Activity Recognition Based on Multiple
Motion Trajectories”, International Conference on Pattern Recognition, pp. 199 -
202, 2004.
[4] Vincent Rabaud, Serge Belongie, “Counting Crowded Moving Objects”,
IEEE Computer Society Conference on Computer Vision and Pattern Recognition, pp.
705 - 711, 2006
[5] G. Brostow and R. Cipolla, "Unsupervised bayesian detection of independent
motion in crowds," IEEE Conference on Computer Vision and Pattern Recognition,
pp. 594-601, 2006.
[6] V. Rabaud and S. Belongie, "Counting crowded moving objects," IEEE
Conference on Computer Vision and Pattern Recognition, pp. 705-711, 2006,
[7] Gavrila, D.M, “Bayesian, Exemplar-Based Approach to Hierarchical Shape
Matching”, IEEE Transactions on Pattern Analysis and Machine Intelligence, pp.
1408 – 1421,2007
[8] Bo Wu, Ram Nevatia, “Detection of Multiple, Partially Occluded Humans in a
Single Image by Bayesian Combination of Edgelet Part Detectors ” , IEEE
International Conference on Computer Vision, pp. 90 - 97 ,2005.
[9] Si-Cheng Zhang, Zhiqiang Liu, “A Robust Real Time. Ellipse Detector”, Pattern
Recognition, vol. 38, pp. 273-287, 2005 .
45
[10] T. Zhao and R. Nevatia, "Bayesian human segmentation in crowded situations,"
IEEE Conference on Computer Vision and Pattern Recognition, pp. 459-266, 2003.
[11] Lu Wang, Nelson H.C. Yung, “ CROWD COUNTING AND
SEGMENTATION IN VISUAL SURVEILLANCE”, International Conference on
Image Processing, pp. 2573 - 2576, 2009
[12] John F. Canny, “A Computational Approach To Edge Detection”, IEEE
Transactions on Pattern Analysis and Machine Intelligence, vol.8, pp. 679-714, 1986
[13] M. Perreira Da Silva, V. Courboulay, A. Prigent, P. Estraillier, “Fast, low
resource, head detection and tracking for interactive applications”, PsychNology
Journal, pp. 243 -264, 2009
[14] Jau Ruen Jen, Mon Chau Shie, and Charlie Chen, “A Circular Hough
Transform Hardware for Industrial Circle Detection Applications”, IEEE Conference
on Industrial Electronics and Applications, pp. 1 - 6, 2006
[15] Qixiang Ye, Jianbin Jiao, Hua Yu, “Multi-posture human detection in video
frames by motion contour matching”, Asian Conference on Computer Vision,
pp.896-904, 2007
[16] Ssu-Wei Chen, Luke K. Wang, Jen-Hong Lan, “Moving Object tracking Based
on Background Subtraction Combined Temporal Difference ” , International
Conference on Emerging Trends in Computer and Image Processing, pp.16-19, 2011
[17] P. Spagnolo, T.D’ Orazio, M. Leo, A. Distante, “Moving object segmentation
by background subtraction and temporal analysis”, Image and Vision Computing,
pp.411-423, 2006
[18] S.E. Umbaugh, “Computer Vision and Image Processing: A Practical Approach
Using CVIPtools”, (Upper Saddle, NJ,Prentice Hall, 1998)
[19] J.F Canny, “A computational approach to edge detection”, IEEE Transactions on
Pattern Analysis and Machine Intelligence, vol. PAMI-8 pp. 679-698, Aug. 1986.
46
[20] J. B. Peter and E H Adelson, “The Laplacian Pyramid as a compact image code”,
IEEE Transaction on Communications, vol. Com-31, No. 4, pp. 532-540, April 1983.
[21] Open Source Computer Vision Library http://www.opencv.org.cn/index.php
[22] Zhe Lin, Larry S. Davis, David Doermann, and Daniel DeMenthon,
“Hierarchical Part-Template Matching for Human Detection and Segmentation”,ICCV,
pp. 1-8, 2007