摘要
在影片中偵測人臉時，可能會遭遇到光線不足或過亮、強光陰影、多個靜止或移動中的物體、物體互相交錯重疊等情況。本研究所發展出一套動態人臉偵測系統，來儘量克服上述各種情況，並且能正確且快速地在影像中偵測到人臉。本系統先用動態續影像交錯相減，準確地找出移動中的物體。之後為了減少光線對影像的影響，先使用YIQ色彩模型來偵測膚色區域，然後再使用HSV色彩模型偵測臉部特徵的位置，也就是顏色基礎(Color-based)和特徵基礎(Feature-based)互相配合使用，進而強化了受到光線或角度改變的人臉特徵，因此增進了人臉偵測的效果。最後本系統使用「2007 IPPR技術競賽－視訊中之人臉偵測技術」所提供一段由實機所拍攝之影片，進行評估與優缺點分析。
並列摘要
There are many difficulties for face detection in video. For examples: the lighting may be insufficient or too bright, many moving objects overlap each other, and so on. This paper proposes a dynamic face detection system to overcome the above difficulties by using a fast approach. At first, the moving regions are found by using image subtraction. To reduce the impact of the light, we use YIQ color model to detect the regions of skins and HSV color model to detect the positions of faces. That is, we combine color-based and feature-based approaches to enhance the features of faces and to avoid the influence of the lighting conditions. At last, the proposed system are evaluated and analyzed by the video from the 2007 IPPR contest.
延伸閱讀
- 周昱呈（2017）。使用多攝影機作人群異常行為偵測〔碩士論文，國立交通大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0030-0705201811562297
- 楊山河（2007）。具物體追蹤與影像重點擷取之人臉監控系統研究〔碩士論文，元智大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0009-2507200700493000
- Chang, E. S. (2010). 部分遮蔽情況下之人形偵測 [master's thesis, National Tsing Hua University]. Airiti Library. https://www.airitilibrary.com/Article/Detail?DocID=U0016-2203201110480777
- 林美亨（2006）。Face Detection in Scalable Video〔碩士論文，國立交通大學〕。華藝線上圖書館。https://doi.org/10.6842/NCTU.2006.00001
- 張元睿（2004）。Face Detection in Still Image〔碩士論文，國立臺灣大學〕。華藝線上圖書館。https://doi.org/10.6342/NTU.2004.02483