臺灣博碩士論文加值系統

生命安全為人類在從事水上活動中最重要的，游泳池都會安排一個救生員在旁待命。到目前為止，市面上還是沒有一套輔助救生員的系統，主要是在溺水辨識上的準確率會有很多問題產生。因此，運用機器視覺的研究方法，來解析其問題的困難點與因應之道，為本研究之主要目的。本研究是經由監視器截取畫面，之後再用群聚演算法、高斯混合模型、模糊隱藏式馬可夫模型等方法，來建立一套完整的溺水辨識系統，進而做到溺水徵兆的警報及保障游泳者的生命安全。經由上述之方法，實作姿態辨識使用游泳者的四個特徵，其為游泳者的橢圓比例、游泳者的面積大小、游泳者的移動速度及游泳者的移動方向，最後在模擬游泳池與真實游泳池作驗證，模擬游泳池只有一個模擬游泳者，得到游泳者偵測率為91%，而游泳者辨識率為88%，真實游泳池有四個游泳者，得到游泳者偵測率為93%，而游泳者辨識率為80%，一旦偵測到溺水情形則發出警報，以減少因救生員沒注意所造成的溺水時間延長及提高存活率。游泳池中存在著許多值得研究之議題，本研究除了針對溺水辨識進行探討外，也可辨識出其他六種游泳行為（自由式、蛙式、蝶式、仰式、仰漂、水中行走），然而運用同樣之方法流程可探究出更多的辨識問題與提出解決方案。

The safety of human life is the most important in water activities, so a lifeguard will be arranged the swimming pool to stand by. So far, due to the lack of a lifeguard auxiliary system in the market, the mainly drowning of the accuracy of identification have a lot of problems. Therefore, the use of machine vision research methods, is to resolve their problems. The difficulties of coping point is the main purpose of this study. This study was intercepted by surveillance aircraft screen, and then after the clustering algorithm, Gaussian mixture model, fuzzy hidden Markov models and other methods to create a complete set of drowning recognition system, and further sent a sign of warning and protection of drowning swimming lives. By the above method to implement gesture recognition, we use the swimmer''s four characteristics, the ratio of the ellipse for the swimmers, swimmer''s size, swimmer''s speed and direction of movement of swimmers, at the last compare in a simulated swimming pool and real for verification. Simulation is only a simulated pool swimmer to get swimmer detection rate of 91%, while the swimmer recognition rate of 88%. The real pool of four swimmers gets swimmers detection rate of 95%, while swimming 80% were identified. Once the situation of drowning is detected, an alarm signal will be sent right away, in order to reduce the possibility that lifeguards fail to notice the result of drowning and prolong survival time. Swimming pools, where many issues are worthy of research, this study discusses not only the recognition for the drowning, but also to identify other six swimming behaviors (freestyle, breaststroke, butterfly, backstroke, backfloat, walking in the water). However, the same method can be used to explore the process of more identification problems and propose solutions.

摘 要 i
ABSTRACT ii
誌 謝 iv
目 錄 v
表目錄 viii
圖目錄 ix
第一章 緒論 1
1.1 研究背景 1
1.2 研究動機與目的 1
1.3 論文貢獻 3
1.4 論文架構 3
第二章 文獻回顧 4
2.1 前言 4
2.2 溺水辨識相關研究 7
第三章 系統架構 9
3.1 前言 9
3.2 背景模型 9
3.3 游泳者偵測 11
3.4 游泳者辨識 13
第四章 背景模型 15
4.1 前言 15
4.2 色彩空間 15
4.3 均值位移演算法 16
4.4 k-平均演算法 18
4.5 期望最大化演算法 18
4.6 高斯混合模型 19
4.7 實作背景模型 20
第五章 游泳者偵測 21
5.1 前言 21
5.2 霍夫轉換 21
5.3 水花去除 24
5.4 形態學 26
5.4.1 膨脹 26
5.4.2 侵蝕 27
5.4.3 斷開 28
5.4.4 閉合 29
5.4.5 連通成份標示法 30
5.5 特徵擷取 31
5.6 實作游泳者偵測 34
第六章 游泳者辨識 35
6.1 前言 35
6.2 模糊理論 35
6.3 隱藏馬可夫模型 36
6.3.1 定義符號說明 36
6.3.2 基本的應用問題 38
6.3.3 基本的應用方法 38
6.3.4 應用方法介紹 40
6.4 特徵轉換 44
6.5 k-最鄰近法 46
6.6 支持向量機 47
6.7 實作游泳者辨識 48
第七章 實驗結果與分析 55
7.1 前言 55
7.2 游泳影片說明 55
7.3 系統流程說明 58
7.4 背景模型結果 59
7.5 游泳者偵測結果 60
7.6 游泳者辨識結果 65
7.7 實驗討論 73
第八章 結論與未來展望 81
8.1 結論 81
8.2 未來的展望 83
參考文獻 84
作者簡介 86

[1] 卓柏全，變動複雜背景下物體追蹤之研究，碩士論文，國立臺北科技大學自動化科技研究所，台北，2009。
[2] H. Uǧuz, H. Kodaz, "Classification of Internal Carotid Artery Doppler Signals Using Fuzzy Discrete Hidden Markov Model," Expert Systems with Applications, vol. 38, no. 6, 2011, pp. 7407-7414.
[3] I.S. Kim, H.S. Choi, K.M. Yi, J.Y. Choi, S.G. Kong, "Intelligent Visual Surveillance - A Survey," International Journal of Control, Automation and Systems, vol. 8, no. 5, 2010, pp. 926-939.
[4] Wen-Hung Liao, Zhung-Xun Liao, Ming-Je Liu, "Swimming Style Classification from Video Sequences," 16th IPPR Conference on Computer Vision, Graphics and Image Processing, Kinmen, ROC, 2003, pp. 226-233.
[5] 陳文輝、卓柏全，「以電腦視覺為基礎之智慧型溺水偵測系統」，第13屆人工智慧與應用研討會，宜蘭，2008，第821~826頁。
[6] H.-L. Chen, M.J. Tsai, C.C. Chan, "A Hidden Markov Model–based Approach for Recognizing Swimmer’s Behaviors in Swimming Pool," International Conference on Machine Learning and Cybernetics, Qingdao, China, 2010, pp. 2459-2465.
[7] W. Lu, Y.-P. Tan, "A Camera-based System for Early Detection of Drowning Incidents," IEEE International Conference on Image Processing, Rochester, New York, 2002, pp. III/445-III/448.
[8] W. Lu, Y.-P. Tan, "A Vision-Based Approach to Early Detection of Drowning Incidents in Swimming Pools," IEEE Transactions on Circuits and Systems for Video Technology, vol. 14, no. 2, 2004, pp. 159-178.
[9] H.-L. Eng, K.-A. Toh, W.-Y. Yau, J. Wang, "DEWS: A Live Visual Surveillance System for Early Drowning Detection at Pool," IEEE Transactions on Circuits and Systems for Video Technology, vol. 18, no. 2, 2008, pp. 196-210.
[10] H.-L. Eng, K.-A. Toh, A.H. Kam, J. Wang, W.-Y. Yau, "An Automatic Drowning Detection Surveillance System for Challenging Outdoor Pool Environments," Proceedings of the IEEE International Conference on Computer Vision, Nice, France, 2003, pp. 532-539.
[11] A.H. Kam, T.K. Ann, E.H. Lung, Y.W. Yun, W. Junxian, "Automated Recognition of Highly Complex Human Behavior," Proceedings - International Conference on Pattern Recognition, Cambridge, England, 2004, pp. 327-330.
[12] D. Comaniciu, P. Meer, "Mean Shift: A Robust Approach Toward Feature Space Analysis," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 24, no. 5, 2002, pp. 603-619.
[13] C. Stauffer, W.E.L. Grimson, "Learning Patterns of Activity Using Real-Time Tracking," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 22, no. 8, 2000, pp. 747-757.
[14] M. Ogaz, R. Sandoval, M. Chacon, "Data Processing from a Laser Range Finder Sensor for the Construction of Geometric Maps of an Indoor Environment," Midwest Symposium on Circuits and Systems, Cancun, Mexico, 2009, pp. 306-313.
[15] 汪惠健譯，論模糊理論與應用，台北：全威圖書有限公司，2006，第21-34頁。
[16] Rabiner, R. Lawrence, "Tutorial on Hidden Markov Models and Selected Applications in Speech Recognition," Proceedings of the IEEE, Stateline, Nevada, 1989, pp. 257-286.
[17] C.-C. Chang, C.-J. Lin, "LIBSVM: A Library for Support Vector Machines," ACM Transactions on Intelligent Systems and Technology, vol. 2, no. 3, 2011, Article 27.