基于监控的视频摘要的研究与实现

发布时间：2018-02-25 22:25

本文关键词： 视频摘要运动目标检测背景建模目标跟踪　出处：《西安电子科技大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着微电子技术和计算机信息技术的飞速发展,视频等海量数据不断积累使得用户对视频的查找和处理越来越困难,这使人们不得不发觉和完善对视频数据的处理的相关技术。如今有关视频处理的视频摘要技术在各行各业发挥着越来越重要的作用,特别是监控方面的应用,大大减少了工作人员的工作量。视频数据的结构是非线性化的,无法按照一般的处理方法对其处理,视频摘要是解决此问题众多方法之一。本文主要研究了视频处理中有关视频摘要的技术,视频摘要的主要过程有读入视频、背景建模、前景提取、运动目标跟踪、目标的时序与空间规划、浓缩视频的生成。针对这几个过程中存在的关键技术进行详细的介绍。一方面在目标检测中的背景模型创建提出了改进算法,另一方面根据改进算法设计了一个视频摘要系统。首先论文研究了运动目标检测算法,根据算法的基本原理将其分为了四类,简单的背景建模、基于统计信息建模、非参数核密度估计建模、非背景建模。对每一类中选择了经典算法进行了介绍和实验。在进行帧差法提取前景的时候提出了一种改进方法,该方法能够有效的减弱粗影问题。其次论文研究了目标跟踪方法,运动目标跟踪是整个设计优化性技术,如果运用的合适得当,可以让系统更加智能人性化,本章主要介绍了卡尔曼滤波,Mean Shift方法。给出了算法的基本原理。论文针对视频摘要最关键的技术,创建背景模型,在平均背景法提出的移动平均背景法的基础上提出了一种基于传统移动平均算法的改进算法,并且对该算法进行了实验分析。最后基于vs2010平台,设计并开发了视频摘要系统,给出了系统的主要模块和流程图,并且给出了实际监控视频测试结果对其进行了相应分析。
[Abstract]:With the rapid development of microelectronics and computer information technology, video and other massive data accumulation makes it more and more difficult for users to find and process video. This has forced people to discover and improve the related technologies for video data processing. Nowadays, video summarization technology about video processing is playing an increasingly important role in various industries, especially in the application of surveillance. The structure of the video data is nonlinear and cannot be processed according to the normal processing method. Video summarization is one of the many methods to solve this problem. This paper mainly studies the technology of video summarization in video processing. The main processes of video summarization include reading video, background modeling, foreground extraction, moving target tracking, etc. The timing and spatial planning of the target, the generation of condensed video. The key technologies in these processes are introduced in detail. On the one hand, an improved algorithm is proposed to create the background model in target detection. On the other hand, a video summary system is designed according to the improved algorithm. Firstly, this paper studies the moving target detection algorithm, and divides it into four categories according to the basic principle of the algorithm, simple background modeling, modeling based on statistical information. Non-parametric kernel density estimation modeling, non-background modeling. The classical algorithms are introduced and experimented in each class. An improved method is proposed to extract the foreground of frame difference method. This method can effectively reduce the coarse shadow problem. Secondly, this paper studies the target tracking method, moving target tracking is the whole design optimization technology, if it is properly used, it can make the system more intelligent and humanized. This chapter mainly introduces the Kalman filter mean Shift method, and gives the basic principle of the algorithm. Based on the moving average background method proposed by the average background method, an improved algorithm based on the traditional moving average algorithm is proposed, and the algorithm is analyzed experimentally. Finally, based on the vs2010 platform, a video summarization system is designed and developed. The main modules and flow charts of the system are given, and the test results of the actual surveillance video are analyzed.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP391.41;TN948.6

【参考文献】