基于商场实时监控视频的云台镜头目标跟踪算法设计与实现

发布时间：2018-04-13 13:14

本文选题：目标跟踪 + 商场云台摄像机　；参考：《山东科技大学》2017年硕士论文

【摘要】：目前,商场的云台镜头监控系统通常是监控人员持续监视屏幕,对进入监控区域的目标通过操作键盘摇杆完成对目标的运动跟踪。本文中的基于商场实时监控视频的云台镜头监控系统选定待跟踪的运动目标,能够根据拍摄环境的变化主动改变观察位置和角度,使视频监控系统能够自主地跟踪选定的目标,并可以放大和缩小场景区域,方便观察运动目标的移动,从而提高了视频监控效率。本文的主要研究内容:首先,介绍视频监控系统的分类,并搭建了本实验所用的云台镜头可变监控系统平台,同时对云台镜头自动跟踪系统进行设计。然后,介绍了 Camshift算法原理,并对传统的Camshift跟踪算法进行仿真。然后对Camshift跟踪算法进行改进,利用颜色特征、HOG特征、LBP纹理特征进行自适应融合,提出了一种多特征自适应融合的Camshift跟踪算法,在每帧视频序列图像中对监控目标进行锁定,记录目标位置,利用多特征进行跟踪,降低了 Camshift跟踪算法光照,遮挡等因素的干扰,提高跟踪算法的稳定性。最后,本论文建立商场云台镜头的自动跟踪模型和镜头自动变倍算法。根据视频画面中监控目标的运动速度和尺寸大小,利用这两个模型自动匹配云台的速度和镜头倍数,对监控目标进行实时跟踪和画面放缩,最终使监控目标始终以合适的尺寸大小显示在视频画面中心区域。并通过程序实现商场云台镜头目标自动跟踪算法以及变倍控制。为了验证本文跟踪算法的有效性,用海康威视摄像头进行了实验仿真。通过海康摄像头录制的视频,对目标进行跟踪,结果显示本文设计的商场云台镜头跟踪系统,系统具有良好的实时性,有稳定的跟踪效果。
[Abstract]:At present, the monitoring system of cloud head lens in shopping mall is usually used to monitor the screen continuously. The target entering the monitoring area is tracked by operating the keyboard rocker to track the motion of the target.In this paper, the shot monitoring system based on real-time surveillance video of shopping mall chooses the moving target to be tracked, which can actively change the observation position and angle according to the change of shooting environment, so that the video surveillance system can track the selected target autonomously.It can enlarge and reduce the scene area, and it is convenient to observe the moving target, thus improving the efficiency of video surveillance.The main research contents of this paper are as follows: firstly, the classification of video surveillance system is introduced, and the platform of cloud head shot variable monitoring system used in this experiment is built. At the same time, the automatic tracking system of cloud head lens is designed.Then, the principle of Camshift algorithm is introduced, and the traditional Camshift tracking algorithm is simulated.Then the Camshift tracking algorithm is improved and the color feature hog feature is used for adaptive fusion. A multi-feature adaptive fusion Camshift tracking algorithm is proposed to lock the monitoring target in each frame of video sequence image.The location of the target is recorded and the multi-feature is used to track the target. The interference of illumination and occlusion of Camshift tracking algorithm is reduced and the stability of the tracking algorithm is improved.Finally, this paper establishes the automatic tracking model and the algorithm of automatic shot variation.According to the moving speed and size of the monitoring target in the video picture, the two models are used to automatically match the velocity of the cloud head and the shot multiple to track the monitored target in real time and the picture is scaled down.Finally, the target is always displayed in the center of the video with the appropriate size.And through the program to realize the automatic tracking algorithm of the shot target and the variable control.In order to verify the effectiveness of the tracking algorithm in this paper, the Hackonview camera is used to carry out the experimental simulation.The video recorded by Haekang camera is used to track the target. The result shows that the system designed in this paper has good real-time performance and stable tracking effect.
【学位授予单位】：山东科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TN948.6

【参考文献】