本篇論文提出一種利用人臉資訊偵測及確認的自動省電節能及安全認證裝置，傳統的消費性電子裝置，當有事離開或忘記關機無人觀看使用時，無自動關機功能，常常浪費電源，並且威脅到安全及保密。而目前最便宜高效率的核能發電，一但污染將嚴重危害人體健康和環境，為響應全球節能減碳及環保概念，利用自動偵測及確認系統，當無人時自動省電，以節省資源並且自動監控。本系統首先是偵測出人臉，再使用主成份分析法(PCA)抽取人臉影像的特徵向量，並同時降低資料維度，將這些特徵向量利用k個最近鄰(KNN)方法，找出最接近的人臉，如是已被核准的本人，即自動啟動裝置。當人有事離開時即自動進入省電模式並且自動監控，如此即可減少電的浪費及增加機器的安全性及保密性。在實驗的部分，本系統的人臉偵測率達93%；在人臉確認率方面本人準確率為93%，非本人準確率為87%。由實驗結果得知，本論文提出以利用人臉資訊偵測的自動省電節能裝置的方法較目前傳統的消費性電子裝置省電，可減少電的浪費，並且提昇電子裝置的安全及保密。In this thesis, an automatic energy saving and security authentication device using real-time face detection and verification is presented. Traditional consumer electronics devices do not have real-time auto shut-off function when users leave or forget to turn off the machine. It consequently results in energy wasting and the threat to security and confidentiality. Although nuclear power is cheap and high-efficiency, it will seriously endanger human health and environment once pollution occurs. To achieve the goals of energy-saving, carbon reduction, and environmental protection and secrecy, we propose a device using real-time face detection and verification to automatically turn off the monitor while human is absent. It will also authenticate the identity of human when the human present in front of the monitor again to determine whether to turn on the monitor or not.In the proposed system, the presented face is detected first and then use Principal Component Analysis (PCA) to extract face features to form face feature vector and decrease the dimensionality of face feature vectors thereafter. Finally, k-th nearest neighbor (KNN) method is employed to find the nearest face candidate. If this face has been authenticated, the monitor will be automatically turn on. If the user leave out of sight, the monitor will be automatically turn off so as to reduce the waste of power consumption while increasing the security and confidentiality of the users.Experimental results demonstrate that 93% accuracy rate of face detection can be attained. The accuracy rate of face verification of user is 93%. The accuracy rate of face verification of non-user is 87%. Experimental results reveal the feasibility and validity of the proposed mechanism in real-time energy saving while enhancing the security and confidentiality of the users.
