面向安防监控的场景变换研究.pdfVIP

万方数据 摘 要 近年来，安防监控系统在交通、军事、机场、银行、商场、社区等场所得到 广泛应用。在诸多领域中，保证安防监控系统正常运行的重要条件之一是获得稳 定、清晰、目标性强的监控视频图像内容。而在实际应用中，因各种因素引起的 场景变换会对监控视频的目标产生影响。为了快速、准确的识别监控视频中出现 的场景变换，本文展开了如下研究： 本文分析了异物遮挡、摄像机偏转、镜头缩放三种典型的场景变换的各自特 点，用模板匹配和边缘能量两种方法检测场景是否发生变换；再分别从颜色信息 和几何信息两个方面进行研究，提出了角点检测法、熵变检测法、子模板匹配法、 中心点扩散寻找特征点的方法分别对几种类型的场景变换进行检测。另外本文还 研究了光线变化对场景变换的干扰，提出了光线变化的识别算法。 在图像一致性检测方面，本文比较了常用的几种彩色图像的颜色向量及其一 致性比较方法，综合了各颜色向量一致性度量标准的优点，提出了改进的颜色自 相关图作为新的实验方法。 最后，本文还收集整理了相关视频和图像库，针对各种不同的场景变换和干 扰及图像的一致性检测进行实验比较。 关键词：场景变换；异物遮挡；摄像机偏转；镜头缩放 I 万方数据 Abstract In recent years, security monitoring systems have been used widely in many places, such as transportation, military, airports, banks, shopping malls, community and so on. In these areas, one of the important conditions to ensure the normal operation of the security monitoring system is to obtain the stable, clear and objective surveillance video content. While in real application, various factors caused the scene obstruction and scene change caused by various factors will affect the target of surveillance video. In order to recognize the scene change timely and timely, this paper carries out the following research: The paper analyzes the characteristics of three typical kinds of scene change, which are object covering, camera displacement and lens scaling, uses two algorithms like model matching method and edge energy method to judge whether the scene have been changed, and then researches the color information and geometry information of the video images, puts