【信息技术】【2017.08】基于差分图像的实时运动检测及MEAN-SHIFT和KALMAN滤波跟踪

本文为美国德克萨斯农工大学（作者：KOOJIN SUNG）的硕士论文，共59页。

计算机视觉的主要技术是检测与跟踪，对监控系统有很大的帮助。本文提出了一种基于单摄像机的实时运动检测与跟踪系统，以降低监控人员的劳动强度。检测算法处理序列帧之间像素的变化，对这些像素值的算术运算提供运动的位置信息，跟踪过程更加复杂。在这个项目中，跟踪系统需要选择感兴趣区域（ROI）作为预处理器。然后，mean-shift算法检测ROI的不同模式，并跟踪每帧的具体模式。为了防止mean-shift跟踪失败，该跟踪系统使用了Kalman滤波器。卡尔曼滤波器利用其位置和速度信息估计、预测期望的均值漂移跟踪路径。该滤波器修正了与路线不可接受的偏差，并帮助跟踪窗口保持正常工作。该项目分别开发了检测算法和跟踪算法，并在最后阶段将它们结合起来。该系统没有采用冗余成像技术，从而使计算时间最小化，最终缩短了实时实现的延迟。该系统将促进低延迟、高性能实时监控系统的发展。

The main techniques of computer vision thatcan be helpful to surveillance system are detection and tracking. This thesisproposes a real-time motion detection and tracking system based on a singlecamera as a cost-effective solution for reducing human labor on surveillance.The detection algorithm deals with the change in pixels between sequentialframes. Arithmetic operations on these pixel values provide positioninformation of motion. Tracking process is more complicated. In this project,the tracking system requires selection of ROI (region of interest) aspreprocessor. Then, mean-shift algorithm examines the distinct pattern of ROIand track the pattern every frame. To prevent a failure of mean-shift tracking,the tracking system is equipped with mathematical tool, Kalman filter. Kalmanfilter estimates and predicts the desirable route of mean-shift tracking, usingits position and velocity information. The filter corrects unacceptabledeviations from the route and helps a tracking window keep functional. Thisproject separately developed detection algorithm and tracking algorithm andcombined them at the final stage. The redundant imaging techniques are excludedin the proposed system in order to minimize the computation time, whichultimately shorten the delay for a real-time implementation. This system willpromote low delay but high performance real-time surveillance system.

1. 引言
2. 本文提出的系统配置
3. 结论与未来工作展望

更多精彩文章请关注公众号：