在即時影音串流中進行人物辨識時，大部分的監控軟體會先擷取人物特徵與資料庫做比對，因為目前的監控軟體擷取的人物輪廓特徵並不完整且需要大量時間與資料庫做比對，所以本論文設計一個快速輪廓擷取(Fast Contour Extraction, FCE)演算法來解決上述的問題，FCE演算法首先將一張畫面分割成數個區塊(Macro Block, MB)，我們假設最左下角的MB沒有包含人物，如果其與任何一個MB的三個顏色分量平均值的差距總和大於一個預設閾值(Threshold)，這些MB就有包含人物。如果有包含人物的MB的所有像素的三個顏色分量均方根差總和大於一個預設閾值，這些MB就有包含人物的輪廓。如果一個有包含人物輪廓的MB，其任何一個像素與最左下角像素的三個顏色分量的均方根差大於一個預設閾值，此像素就是人物的像素。如果一個人物像素周圍的八個像素有任何一個不是人物像素，此人物像素就是輪廓的像素。接著FCE演算法會將擷取的輪廓像素依據一個輪廓的MB數量、像素數量、移動速度、長與寬做人或物的分類。我們在影像辨識伺服器上藉著修改OpenCV的輪廓擷取函式並且新增輪廓分類函式來實作FCE演算法，在實作中我們量測人物輪廓擷取的時間、人物輪廓分類的時間、人物輪廓分類的準確率，我們使用這三項量測數據與YOLOv7做比較，實驗結果可以驗證FCE演算法比起YOLOv7能更快的進行輪廓擷取與人物分類。

In real-time video streaming, figure identification often involves outlining figure features before comparing them to a database. However, existing monitoring software’s faces limitations in capturing complete figure contours and suffers from lengthy processing times evaluated from database comparisons. To address these challenges, we propose the Fast Contour Extraction (FCE) algorithm. The FCE algorithm segments a frame into several Macro Blocks (MBs). It assumes that the MB located in the bottom left corner does not contain the figure. If the total difference between a given MB and the average of the three-color components exceeds a predefined threshold, it indicates the presence of the figure outline. Similarly, if the total root-mean-square deviation of the three-color components in all pixels of the figure-included MB exceeds a predefined threshold, then the corresponding MBs are included in the outline of the figure. The given pixel is one of the figure-included pixels if the total difference of root-mean-square deviation of the three-color components between the pixel on the bottom left corner and any pixel from the figure-included MB exceeds the threshold. Then, if any of the eight pixels surrounding a figure-included pixel does not contain the figure, it is considered as part of the figure contour. Furthermore, the FCE algorithm classifies the captured pixels, including the contour, based on the MB number of contours, total pixels, motion velocity, contour length, and width. We modified the contour capture function from OpenCV on the video identification server and added a contour classification function to implement the FCE algorithm. We evaluated the time required for capturing figure contours, contour classification, and the accuracy of contour classification. These measurements were then compared to those obtained using YOLOv7. Our results demonstrate that the FCE algorithm outperforms YOLOv7 in contour capturing speed and figure classification accuracy.

論文審定書 i
致謝 ii
摘要 iii
Abstract iv
目錄 v
圖目錄 vii
表目錄 viii
第一章 導論 1
1.1 研究動機 1
1.2 研究方法 2
1.3 章節介紹 3
第二章 影音串流的人物輪廓擷取 4
2.1 即時的影音串流 4
2.1.1 RTSP的連線建立 4
2.1.2 連續畫面的壓縮 6
2.2 人物的輪廓擷取與分類 8
2.3 YOLOv7與OpenCV的輪廓擷取 9
2.3.1 YOLOv7的輪廓擷取 9
2.3.2 OpenCV的輪廓擷取 10
2.4 相關研究 12
第三章　人物輪廓的快速擷取 15
3.1 靜態畫面的輪廓擷取 15
3.1.1人物輪廓的擷取 15
3.1.2輪廓擷取的方法 17
3.2 連續畫面的輪廓擷取 22
3.3 人物輪廓擷取的演算法 25
3.3.1 MB與人物 25
3.3.2 MB中的人物輪廓擷取 26
3.4 人物輪廓的分類 27
3.4.1 計算人物輪廓的MB數量 27
3.4.2 計算人物輪廓的像素總數 28
3.4.3 計算人物的移動速度 28
3.4.4 計算人物輪廓的長與寬 29
3.5 人物輪廓分類演算法 30
第四章 實作與結果分析 32
4.1 快速輪廓擷取的實作 32
4.2 FCE與人物分類的虛擬碼 34
4.2.1 MB與人物像素 34
4.2.2 輪廓擷取的函式 36
4.2.3 連續畫面的人物輪廓 39
4.2.4 人物輪廓的分類 42
4.3 實作的結果與分析 44
4.3.1 人物輪廓擷取的時間 46
4.3.2 入物分類的時間與準確率 49
第五章 結論與未來工作 53
5.1 結論 53
5.2 本論文遭遇的困難 54
5.3 未來工作 54
Reference 55
附錄 FCE 演算法的輪廓擷取範例 58
Acronyms 62
Index 63

[1] W. Zheng, I. Ahmad, and M. L. Liou, “Real-Time Software Based MPEG-4 Video Encoder,” Proceedings of Workshop and Exhibition on MPEG-4 (Cat. No.01EX511), San Jose, CA, USA, pp. 71-74, Jun. 2001.
[2] A. Banitalebi-Dehkordi, M. Azimi, M. T. Pourazad, and P. Nasiopoulos, “Compression of High Dynamic Range Video Using the HEVC and H.264/AVC Standards,” 10th International Conference on Heterogeneous Networking for Quality, Reliability, Security and Robustness, Rhodes, Greece, pp. 8-12, Aug. 2014.
[3] H. Schulzrinne, A. Rao, and R. Lanphier, “Real-Time Streaming Protocol (RTSP),” RFC2326, Apr. 1998.
[4] Z. Xu, Y. Fu, and A. Cui, “Stacked Mosaic Detection Based on Line Expansion,” 2023 International Conference on Intelligent Media, Big Data and Knowledge Mining (IMBDKM), Changsha, China, pp. 1-5, Mar. 2023.
[5] A. Vetro, T. Wiegand, and G. J. Sullivan, “Overview of the Stereo and Multiview Video Coding Extensions of the H.264/MPEG-4 AVC Standard,” Proceedings of the IEEE, vol. 99, no. 4, pp. 626-642, Apr. 2011.
[6] B. Ahirwal, M. Khadtare, and R. Mehta, “FPGA Based System for Color Space Transformation RGB to YIQ and YCbCr,” 2007 International Conference on Intelligent and Advanced Systems, Kuala Lumpur, Malaysia, pp. 1345-1349, Nov. 2007.
[7] S. Midha, R. Vijay, and S. Kumari, “Analysis of RGB and YCbCr Color Spaces Using Wavelet Transform,” 2014 IEEE International Advance Computing Conference (IACC), Gurgaon, India, pp. 1004-1007, Feb. 2014.

[8] G. Choi, P. Heo, and H. Park, “Triple-Frame-Based Bi-Directional Motion Estimation for Motion-Compensated Frame Interpolation,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 29, no. 5, pp. 1251-1258, May 2019.
[9] O. D. Trier and A. K. Jain, “Goal-Directed Evaluation of Binarization Methods,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 17, no. 12, pp. 1191-1201, Dec. 1995.
[10] T. R. Konala, A. Nammi, and D. S. Tella, “Analysis of Live Video Object Detection Using YOLOv5 and YOLOv7,” 2023 4th International Conference for Emerging Technology (INCET), Belgaum, India, pp. 1-6, May 2023.
[11] J. Sigut, M. Castro, R. Arnay, and M. Sigut, “OpenCV Basics: A Mobile Application to Support the Teaching of Computer Vision Concepts,” IEEE Transactions on Education, vol. 63, no. 4, pp. 328-335, Nov. 2020.
[12] S. Suzuki and K. Abe, “New Fusion Operations for Digitized Binary Images and Their Applications,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. PAMI-7, no. 6, pp. 638-651, Nov. 1985.
[13] J. Yuan, W. Cao, Z. Cai, and B. Su, “An Underwater Image Vision Enhancement Algorithm Based on Contour Bougie Morphology,” IEEE Transactions on Geoscience and Remote Sensing, vol. 59, no. 10, pp. 8117-8128, Oct. 2021.
[14] R. -W. Bello, A. S. A. Mohamed, and A. Z. Talib, “Contour Extraction of Individual Cattle from an Image Using Enhanced Mask R-CNN Instance Segmentation Method,” IEEE Access, vol. 9, pp. 56984-57000, 2021.
[15] A. Niaz, A. Joshi, A. Munir, D. D. Kim, H. C. Song, and K. N. Choi, “Hybrid Active Contour Based on Local and Global Statistics Parameterized by Weight Coefficients for Inhomogeneous Image Segmentation,” IEEE Access, vol. 8, pp. 57348-57362, 2020.
[16] Y. Hu, “Image Segmentation Based on Velocity Feature Vector for Moving Target Extraction,” IEEE Sensors Journal, vol. 20, no. 20, pp. 11983-11991, Oct. 2020.
[17] S. Wolf, A. Bailoni, C. Pape, N. Rahaman, A. Kreshuk, U. Köthe, and F. A. Hamprecht, “The Mutex Watershed and Its Objective: Efficient, Parameter-Free Graph Partitioning,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 43, no. 10, pp. 3724-3738, Oct. 2021.
[18] M. Zhang, X. Liu, D. Xu, Z. Cao, and J. Yu, “Vision-Based Target-Following Guider for Mobile Robot,” IEEE Transactions on Industrial Electronics, vol. 66, no. 12, pp. 9360-9371, Dec. 2019.
[19] S. Pei, L. Li, L. Ye, and Y. Dong, “A Tensor Foreground-Background Separation Algorithm Based on Dynamic Dictionary Update and Active Contour Detection,” IEEE Access, vol. 8, pp. 88259-88272, 2020.
[20] N. Ghode and P. H. Bhagat, “Motion Detection Using Continuous Frame Difference and Contour Based Tracking,” 2019 3rd International Conference on Trends in Electronics and Informatics (ICOEI), Tirunelveli, India, 2019, pp. 490-493, Apr. 2019.