分散式即時互動Client/Server影像辨識系統

Abstract

近年由於電腦科技的進步，使得視訊監控系統廣泛的佈建在生活周遭。然而在現今的監控系統平台架構下，因系統的運作機制並非相當完善，如欲達到監控的目的，需要大量的人力及時間。以目前的系統架構而言，取像設備、監控中心、監控人員三者大都位於同一區域，也就是說監控人員需要在監控中心存取影像檔案。另外為了判斷是否有異常現象發生，監控人員需要不間斷的觀看相機影像，當有事件發生時，再以人工設定的方式進行錄影的動作。另外當系統提供多個相機影像時，觀看相機的選取需透過人工設定的方式達成。由上可知整個系統架構所提供的人機互動性並不高，往往需要人工設定才能執行相關的動作。 有鑑於監控平台運作方式的不完善，因此我們提出即時互動式監控系統來加以解決。在即時互動式監控系統架構下，存在著Client/Server的概念，Server端負責影像資料的收集，監控人員可透過Client端的程式向Server端存取監控影像，達到遠端檔案存取的功能。另外因為我們在Client端加入物件定位及手勢辨識等影像處理，使Client可對監控影像進行影像處理的動作，進而將運算結果回報給Server端，Server端再依Client端的回報訊息進行相對應的處理。Client/Server的分散式架構搭配相關的影像處理，使系統達到自動錄影機制、自動相機切換機制、遠端檔案存取功能。

Due to the dramatic advances of the computer technology, surveillance systems are widely accepted and used. However, a huge amount of human resources and time are required to monitor these systems due to the lack of a completely operational mechanism. Specifically, for the current model, capture devices, the surveillance center, and the surveillance personnel have to be at the same location to accomplish the task. In other words, the surveillance personnel must always stay in the surveillance center to look at the monitors closely in order to see whether an abnormal event occurs or not. When an abnormal event occurs, the surveillance personnel have to manually turn on the recording devices. If multiple cameras are present, the surveillance personnel also have to manually switch between different cameras. From the above, it is not hard to conclude that the level of human-machine interaction for this model is not high enough. It requires lots of manual settings. To overcome this major drawback, in this paper, a real-time interactive surveillance system is proposed. In this model, the client/server concept is put in practice. The server system is responsible for collecting visual data from various types of cameras. Thus, the surveillance personnel can remotely acquire these collected data with the use of the client system. Moreover, because the object positioning and the gesture recognition techniques are added to the client system, the client system is able to process the acquired image data. When detecting an abnormal event, the client can inform the server about it. The server in turn can react to the event according to the pre-defined rules. The proposed distributed client/server architectures in conjunction with the image processing capabilities provide the surveillance system with the abilities to automatically record, switch between cameras, and access files remotely.