目标跟踪技术在智能视频监控系统中的应用研究

发布时间：2018-03-03 13:31

本文选题：智能监控系统　切入点：目标跟踪　出处：《兰州理工大学》2014年硕士论文　论文类型：学位论文

【摘要】：以计算机视觉技术为基础的智能视频监控系统,目前已经广泛应用于人们生产生活的各个方面。智能监控是集成了智能行为识别算法,能够对画面场景中的人或车辆的行为进行识别、判断,并在适当的条件下,产生报警提示用户；运动目标检测与跟踪是智能视频监控系统中的关键技术之一。本文主要研究了智能视频监控系统中目标跟踪技术的实现方法,其主要研究工作如下： 1.首先对传统智能视频监控系统中常用的3种目标跟踪算法(卡尔曼滤波算法、均值漂移算法、自适应均值漂移算法)进行了研究；然后,通过分别在不同遮挡情况下(无遮挡、半遮挡、全遮挡等),对3种目标跟踪算法进行了仿真实验；并通过对比分析,归纳出了3种跟踪算法的优点以及存在的问题。 2.针对自适应均值漂移算法在目标发生遮挡时鲁棒性差的问题,提出了一种基于自适应均值漂移的改进的目标跟踪算法。基于卡尔曼滤波算法有较好的预测特性以及自适应均值漂移算法具有较高的实时性的基础上,本文将卡尔曼滤波算法应用于自适应均值漂移算法中,来解决自适应均值算法当目标发生遮挡时跟踪失败的问题；在目标跟踪过程中通过巴氏系数来判断目标是否被遮挡,当目标出现遮挡时运用卡尔曼滤波进行预测,然后把预测的结果作为自适应均值漂移算法的下一次输入,在目标出现对其进行重新快速的捕获。仿真实验结果证明,本文所提出的改进算法具有较好的实时性,且在目标遮挡情况下具有较强的鲁棒性。 3.在Microsoft Visual Studio2008集成开发环境下,采用QT应用程序框架及OpenCV计算机视觉库代码,实现了基于运动目标检测与跟踪的视频监控系统。该系统对于USB摄像头或AVI视频文件输入的视频,能实时检测出场景中的运动物体并进行跟踪。
[Abstract]:Intelligent video monitoring system based on computer vision technology, has been widely used in all aspects of people's life and production. The intelligent monitoring system integrated intelligent behavior recognition algorithms to picture the scene in person or vehicle behavior recognition, judgment, and under appropriate conditions, an alarm prompts the user; moving target detection and tracking is one of the key technologies in intelligent video surveillance system. This paper mainly studies the method to realize the target tracking technology in intelligent video surveillance system, the main research work is as follows:
1. the first of the 3 objectives of the traditional intelligent video surveillance system tracking algorithm (Calman filtering algorithm, mean shift algorithm, adaptive mean shift algorithm) is studied; then, through respectively in different occlusion condition (no occlusion, half occlusion, occlusion, etc.) of 3 kinds of target tracking algorithm experiment; and through comparative analysis, summed up the advantages of the 3 kinds of tracking algorithm and existing problems.
2. adaptive mean shift algorithm in target robust occlusion problems, put forward a kind of improved adaptive mean shift tracking algorithm based on target. Calman filtering algorithm based on prediction and better performance of adaptive mean shift algorithm has high real-time on the basis of the Calman filter algorithm applied to adaptive mean shift algorithm, to solve the adaptive k-means algorithm when the target is occluded by tracking failure problem; Bhattacharyya coefficient to judge whether the target is occluded in the target tracking process, when the target when there is occlusion is predicted using the Calman filter, and then the forecast result as the adaptive mean shift algorithm for the next input in the target there re quick to capture it. The simulation results show that the algorithm proposed in this paper has better It is real-time and has strong robustness in the case of target occlusion.
3. in the Microsoft Visual Studio2008 integrated development environment, using QT application framework and OpenCV computer vision library code, the realization of the video monitoring system for moving target detection and tracking based on the input system for the USB camera or AVI video video files, can be used to detect moving objects in the scene and track.

【学位授予单位】：兰州理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN948.6

【参考文献】