面向监控视频的行人跟踪算法设计与实现
关键词 视频监控，行人跟踪，实时应用
毕业论文设计说明书外文摘要
Title Design and Implementation of Pedestrian Tracking for Video Surveillance
Abstract
The purpose of video surveillance is to detect, track and understand the target motion, utilizing video images captured by cameras. As the rapid development of computer, communication and electronic technology, video surveillance technology has been widely researched and applied, gradually becoming a hot research in the field of computer vision. Pedestrian tracking is the core content of the video surveillance system. In recent years, domestic and foreign researchers have proposed some relevant pedestrian tracking algorithms. This paper introduces seven advanced real-time tracking algorithms. They are TLD, struck, mean shift, LSST, CT, L1-APG, IVT and CT. Based on the in-depth study of TLD algorithm, experiments and analysis are implemented, and the experiment results are compared with those of struck, L1-APG algorithm. Aiming at solving the problems of TLD algorithm, an improved method is proposed.
Keywords video surveillance, pedestrian tracking, real-time application
目 次
1 绪论 1
1.1 引言… 1
1.2 课题背景及研究意义 1
1.3 国内外研究现状 2
1.4 主要研究内容… 6
1.5 论文的结构安排 6
2 目标跟踪算法介绍 7
2.1 引言… 7
2.2 TLD算法 7
2.3 Struck算法 8
2.4 Mean Shift算法… 8
2.5 LSST算法… 9
2.6 L1-APG算法 10
2.7 IVT算法… 10
2.8 CT算法 11
2.9 本章小结… 11
3 TLD算法研究与实现… 12
3.1 引言… 12
3.2 TLD框架… 12
3.3 TLD流程图 14
3.4 TLD实验与分析 15
3.5 本章小结… 25
4 未来工作展望 26
4.1 引言… 26
4.2 基于SIFT和HOG的粒子滤波器 26
4.3 TLD算法的改进方案 27
4.4 本章小结… 28
结论 … 29
致谢 … 30
参考文献… 31
图1.1 struck支持向量集的可视化 8
图3.1 TLD框架图… 12
图3.2 检测模块流程图… 14
图3.3 学习模块流程图… 14
图3.4 TLD流程图… 15
图3.5 TLD数据集… 17
图3.6 PKU-SVD-B数据集 18
图3.7 01_David数据中的五帧输出结果 20
图3.8 04_pedestrian2数据中的五帧输出结果… 20
图3.9 07_motocross数据中的五帧输出结果 21
图3.10 10_panda数据中的五帧输出结果… 21
图3.11 imgs4数据中的四帧输出结果… 23
图3.12 imgs2数据中的四帧输出结果… 24
图3.13 imgs5数据中的四帧输出结果… 24
图3.14 imgs9数据中的四帧输出结果… 25
表3.1 TLD数据集说明… 16
上一篇：组合聚类算法研究与实验+文献综述