應用資訊隱藏技術做安全監控之研究

Abstract

現今，越來越多監控視訊被使用來以預防犯罪。本論文提出三種資訊隱藏技術在視訊監控系統上的應用。第一個應用是在視訊中尋找特定的人和移動物體的方法。我們抽取出視訊影像中移動部分的特徵並且將其隱藏在影像之中，之後藉此來尋找具有相同移動物體特徵的影像。第二個應用是把含有移動資訊的影像保存在視訊中較安全的片段中，以預防被惡意竄改。我們將含有移動資訊的可疑影像隱藏到背景場景中的I框架影像(I-frame)中，並將預先定義好的驗證訊號藏入其中。經由一本文所提驗證程序，在視訊中修改或置換相關的影像可被偵測出來，並且可經由還原程序將影像還原回來。最後，第三個應用在監視汽車被非法入侵的事件。在車上的電腦會將警告訊息以及車牌號碼經由無線網路傳送到監視系統，這些訊息接著會被即時隱藏到相關的視訊影像中，因此入侵事件的影像可以被檢索出來。最後我們提出了相關的實驗結果，證明了所提方法之可行性。

Nowadays, more and more surveillance videos are recorded to prevent crimes. Three methods for data hiding applications implemented on video surveillance systems are proposed. In the first method proposed for the purpose of searching videos for specified persons or motion objects, we extract features from motion parts of video frames and embed them in the frames to help later searches for frames which contain motion objects with similar features. In the second method proposed for keeping motion frames in safer video sections to prevent malicious tampering, we embed concerned motion frames together with specially-designed authentication signals into I-frames with background scenes. Modification or replacement of relevant motion frames can be checked via an authentication process and compensated by a frame recovery process. Finally, the third method is proposed to monitor car break-in events, in which a warning message of event time and the car plate number is sent wirelessly from an on-car computer to a surveillance system. The message is then embedded in relevant video frames. Such event frames can be retrieved later for inspection. Good experimental results prove the feasibility of the proposed methods.