摘要随着计算机技术的不断发展,视频跟踪成为一个热点问题。其中,多摄像机下的目标跟踪因其优越性受到越来越多研究人员的关注。多摄像机的跟踪问题主要涉及以下两个问题:一是单摄像机的跟踪实现;二是多摄像机间的交接。63012
本文首先介绍了目标检测与跟踪的原理,具体以均值偏移算法结合卡尔曼滤波的方法实现单摄像机下的目标跟踪。进而讨论了多摄像机跟踪系统,包括系统结构、系统同步等。重点讨论了多摄像机间的目标交接问题,首先阐述了相关原理,之后给出了一种基于SURF特征的解决方案。最后设计了双摄像机的目标跟踪系统,并给出相应的实验结果。
毕业论文关键词 目标检测与跟踪;均值漂移;卡尔曼滤波;多摄像机跟踪;目标交接;SURF特征匹配
毕业设计说明书(论文)外文摘要
Title Design of Logistics Tracking System Based on Multi-Camera
Abstract With the continuous development of computer techniques, video tracking is becoming a hotspot. In this field, object tracking based on multi-camera catch eyes of increasing researchers due to its advantages. The topic involves two main problems: 1)how to realize the object tracking based on a single camera; 2)how to connect one camera with another to keep on tracking.
The principles of object detection and tracking are introduced first in this paper. Then I elaborate a scheme, which combines algorithm of Mean Shift with Kalman filters, that can achieve real-time object tracking with a single camera.In the next sections, there are discussions on multi-camera tracking system, including its structure and the synchronization. This paper also focuses on the effective connections between different cameras. After explaining the relevant principles, I provide a scheme on the base of SURF feature in order to solve the problem of object handoff. At last, the design of object tracking system based on dual-camera is given with corresponding experiment results.
Keywords object detection and tracking, Mean Shift, Kalman Filters, Multi-camera tracking, object handoff, SURF feature matching
1 绪论 1
1.1 课题研究背景及意义 1
1.2.1 国外研究现状 2
1.2.2 国内研究现状 3
1.3 本文主要工作及论文安排 3
2 运动目标的检测与跟踪 5
2.1 运动目标检测 5
2.1.1 帧间差分法(时间差分法) 5
2.1.2 背景减除法 6
2.1.3 光流法 7
2.1.4 几种算法的比较 7
2.2 运动目标跟踪 8
2.2.1 基于区域的跟踪[1] 8
2.2.2 基于特征的跟踪 9
2.2.3 基于变形模板的跟踪 9
2.2.4 基于模型的跟踪 9
3 基于Mean Shift的单摄像机目标跟踪 11
3.1 基于Mean Shift的跟踪算法 11
3.1.1 目标模型