智能监控下的运动检测与跟踪

发布时间：2019-03-29 19:14

【摘要】：智能监控技术在交通、金融、安全等部门有着广泛的应用,而且随着社会的发展,智能监控的发展也越来越重要;它是计算机视觉研究范围的一个重要内容,而运动目标检测跟踪则是智能监控系统的一个核心技术。运动目标检测研究的对象是视频的图像序列,其目的是判断视频序列中是否存在运动对象,运动目标跟踪主要指追踪检测目标的运动轨迹,及运动的相关参数。本文主要以智能监控中的运动目标的跟踪检测为研究内容,为后续的智能监控发展提供基础。在运动目标检测方面,本文结合了帧间差分法和背景差分法的优缺点,提出了帧差背景相结合的方法,实验验证该方法能够提高检测的精准度,减少误差。对于运动目标跟踪,本文采用了基于Mean-shift算法的Cam-shift算法,该算法通过自动调节搜索窗口大小来适应目标的大小的方式对Mean-shift算法进行了相应的改进。但是Cam-shift算法自身还是会产生误差,它本身能够解决遮挡问题,但是由于帧间位移较大时,会产生误差,出现目标跟踪丢失的情况,在改进Cam-shift方面本文提出事先对运动目标进行简单的运动预估,通过估算值平移搜索框,增大搜索框与目标联接的可能性,提高跟踪的准确性。
[Abstract]:Intelligent monitoring technology is widely used in transportation, finance, security and other departments, and with the development of society, the development of intelligent monitoring is becoming more and more important. It is an important part of computer vision research, and moving object detection and tracking is a core technology of intelligent monitoring system. The object of moving object detection is the image sequence of video. Its purpose is to judge whether there are moving objects in the video sequence. Moving target tracking mainly refers to tracking the moving track of detected target and the related parameters of motion. This paper mainly focuses on the tracking and detection of moving objects in intelligent monitoring, which provides the basis for the development of intelligent monitoring. In the aspect of moving target detection, this paper combines the advantages and disadvantages of inter-frame difference method and background difference method, and puts forward a method of combining frame difference background. The experimental results show that this method can improve the precision of detection and reduce the error. For moving target tracking, this paper adopts the Cam-shift algorithm based on Mean-shift algorithm, which improves the Mean-shift algorithm by automatically adjusting the size of the search window to adapt to the target size. However, Cam-shift algorithm itself will produce errors, it can solve the occlusion problem itself, but because of the large inter-frame displacement, it will produce errors, resulting in the loss of target tracking. In the aspect of improving Cam-shift, this paper proposes a simple motion prediction method for moving objects in advance. By estimating the translation search box, the possibility of linking the search box with the target is increased, and the accuracy of tracking is improved.
【学位授予单位】：东南大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】