【作者】 黄孝建；

【导师】 杨东勇；

【作者基本信息】 浙江工业大学 ， 计算机软件与理论， 2011， 硕士

【摘要】 物体遗留和物体移走检测、绊线检测是视频监控中的重要技术。本文综述了国内外物体遗留和物体移走检测、绊线检测的研究现状,设计实现了新的物体遗留和物体移走检测、绊线检测方法。论文完成的主要工作如下:针对物体遗留和物体移走检测,采用统计模型建立短期背景和长期背景,背景差和相邻帧差相结合提取运动目标,对短期背景非运动区域进行实时更新。每隔? 10?帧将短期背景和长期背景差分,提取目标区域,用边缘特征判定法和巴氏距离判定法相结合的方法识别目标区域是物体遗留还是物体移走。设计开发了具有物体遗留和物体移走检测的监控系统原型,设置不同复杂程度、不同背景的场景,以箱子和包裹作为遗留物体或移走物体,对所设计的物体遗留和物体移走检测方法进行实验验证,结果表明,所设计的方法能够准确实现物体遗留和物体移走检测。在绊线检测中,采用高斯模型进行背景建模,采用背景差分提取前景目标,扫描绊线上所有点,当检测到目标穿越绊线时,若为双向绊线,系统直接报警,若为单向绊线,利用目标颜色直方图欧拉距离来确定目标运动方向,如果目标运动方向与禁止穿越方向一致,系统报警。设计开发了具有绊线检测的监控系统原型,选择不同的实验场景,设置不同位置和类型的绊线,对所设计的绊线检测方法进行实验验证,结果表明,所设计的方法能够准确实现绊线检测。更多还原

【Abstract】 Abandoned object detection, removed object detection and trip-line detection are important techniques in video surveillance. Based on the summary of present research situation, novel methods of abandoned object detection, removed object detection and trip-line detection are proposed and testified in this paper. The main works are as follows:For abandoned object detection and removed object detection, the statistical model is used to establish short-term background and long-term background. Background differencing and coterminous frame differencing are combined to obtain the moving object, and only the non-moving region of short-term background is updated in real-time. The object region is detected by differencing the short-term background and long-term background each 10 frames. When the changing region is detected, edge determinant method and Bhattacharya coefficient determinant method are combined to distinguish whether the changing region is abandoned object or removed object. A surveillance system composed of abandoned object detection and removed object detection is designed and developed in this paper. Different backgrounds are set up by selecting different scenes and artificially interposing jam, where various boxes and packages are selected as abandoned or removed objects. The given abandoned object detection and removed object detection method is verified in several areas. Finally, the experiments testify that the proposed method can detect abandoned and removed objects accurately.For trip-line detection,multi-Gauss Model is used to model background. Moving objects are detected by background differencing. When foreground object is detected on trip-lines, for double-direction trip-line, system directly alerts while for single-direction trip-line, the moving direction of the object must be analyzed at first. If the moving direction of the object matches forbidden passing rule, system alerts. A surveillance system of trip-line detection is designed and developed. Different backgrounds are set up by selecting different scenes different trip-lines are defined on the scenes, the given trip-line detection method is verified in several areas. The experiments prove the accuracy of the proposed method for trip-line detection.更多还原