人臉辨識應用，經常都在高性能主機的伺服器上執行，但在雲端伺服器上，無法滿足即時性和可靠性需求。 本研究提出一個智慧型視訊監控物聯網架構，對此，應用裝置端是低硬體資源的攝影機，實作嵌入式人臉偵測；閘道器端是邊緣計算(edge computing)伺服器，執行深度學習人臉辨識。而雲端服務是一個Web界面，監控人機界面，使用者可由遠端，進行設定與監控目前影像狀況。 實驗驗證結果顯示，攝影機端能快速偵測，並擷取人臉影像，透過物聯網，傳至閘道器，執行深度學習，實現高效率人臉辨識。一來，滿足即時人臉辨識的需求，二來，減少網路頻寬的負擔。透過此研究，提高視訊監控物聯網系統整體效能，進而使物聯網時代快速推動。 ;Face recognition application can execute in the server of high-performance mainframe constantly, yet it cannot meet the requirements of instantaneity and reliability in the cloud server. This thesis aims to investigate the Internet of Things (IoT) framework toward intelligent video surveillance (IVS). The device end is a hardware-constrained camera embedding with face detection system whereas the gateway end is an edge computing server, executing deep learning face recognition. Moreover, cloud service is a web interface embedding with human machine interface, thus the user can remote the setting and monitoring for the current image status. The findings of this research shows that the camera end can quickly detect and capture the face image to the gateway end in virtue of the IoT. It executes the deep learning and achieves high efficiency technology regarding face recognition. For one thing, IVS can meet the demand of real time face recognition. For the other, it can lighten the burden of network bandwidth. I believe that the findings from my thesis can elevate the overall effectiveness of IVS, and then open up a new era for the IoT.
