视频监控中运动图像检测与识别技术的研究

【作者】 范莹；

【导师】 郭成安；

【作者基本信息】 大连理工大学 ， 信号与信息处理， 2003， 硕士

【摘要】 场景监视技术是现代IT技术中的一类重要应用领域。场景监视技术不仅仅在社会日常生活和工作中有着非常广泛的应用前景，在能源，国防，交通等领域也正在发挥着日益重要的作用。此外，场景监视技术中的核心技术是运动目标的检测、提取和识别，这也是目前正在热门研究的视频压缩和传输中的关键技术之一。 本文基于安全智能报警的应用背景，通过运用数字图像和模式识别理论，设计了一套视频监控中运动图像检测与识别算法。本文重点研究了运动检测、运动目标的恢复和提取以及运动目标的识别三部分。在运动检测中，根据CCD器件成像特征，建立了一种高斯运动变化检测模型，检测出视频序列图像中的运动变化区域。在运动目标的恢复与提取中，本文提出了两种算法：一种是两帧差分法恢复与提取运动目标，另一种是三帧差分法提取和恢复运动目标。而运动物体的识别则是基于安全运动报警系统设置的模糊判决算法，其中为了提取有效的判决参数，本文在图像空域分割中还提出了一种顺序区域生长法，可在一次性图像扫描过程中实现像素连通判别和特征参数提取。对于上述各个算法，本文均进行了模拟实验，结果表明，综合运用本文中提出的图像处理和识别方法，可以正确、快速检测和识别视频序列中的运动目标。更多还原

【Abstract】 Scene monitoring technology is one of the chief fields in IT technology. Scene monitoring technology not only has a promising broad future in human’s day life but also plays a more and more important role in many fields such as resource, national defense, traffic and etc, among which intelligent safe guarantee system is mostly used in men day life. Furthermore, the technology for detection and recognition of moving objects in images, the core technology in the scene monitoring system, is else one of the making technologies in thefield of video compression and transmission.According to the applied background of intelligent safe guarantee system, this paper develops digital image processing and pattern recognition methods and designs a method to detect and recognize moving objects in video sequences. This paper puts emphasis on the moving and change detection, extraction of moving objects and moving objection recognition. In moving and change detection, a Guass moving and change detection model is designed based on the imaging character of CCD in order to detect moving and change area in sequential images. In moving objects extraction this paper this paper proposed two methods, one is two image difference method to extract moving objects, another is three images difference method. Moving object recognition adapts fuzzy method based on the applied background of intelligent safe guarantee system. In order to successfully implement the space domain segmentation, this paper also develops an algorithm that can distinguish connexity between pixels and extract the feature parameters in one image scanning procedure. Simulation experiments have been made. And the results show that, moving objects in sequential images can be detected and recognized correctly by using the algorithm in this paper.更多还原