摘要
智能监控系统中的行人跟踪功能在很大程度上能够减轻工作人员的大量眼力工作，通过智能的方式自主地跟踪用户感兴趣的目标。在视频监控环境的基础上，利用改进的粒子滤波器，设计了一个行人跟踪系统，并使其具有系统性能分析的功能。通过使用Mean Shift算法进行目标位置的预估计和多特征融合的方式使该系统在视频背景复杂的情况下能够实现对行人的稳定跟踪。并在实际应用中，分析了在遮挡情况下，行人跟踪效果差的原因，并予以改进。在Visualc＋＋集成开发环境中，基于OpenCV和MATLAB编程实现了系统，能够实时地对目标进行跟踪，并对跟踪结果进行反馈。
The functionality of pedestrian tracking in intelligence supervisory system can automatically track the interesting target, needing little human effort. Based on the video surveillance system, this paper designs and implements a pedestrian tracking system which leverages the improved Particle Filter method to track pedestrian. In addition, the system could analyze and evaluate the tracking performance. It leverages the approach of pre-es- timation and multi-feature fusion by means of Mean Shift to achieve a stable pedestrian tracking in the complex backgroud.This article also analyzes the cause of poor tracking performance in the case of shelter, and improves it. This system is implemented in IDE Visual C＋＋, with OpenCV and MATLAB. Experiment results show that it can carry on feedback on the tracking result, and its tracking performance achieves the real-time level.
作者
朱然
马培军
苏小红
ZHU Ran, MA Peijun, SU Xiaohong （School of Computer Science and Technology, Harbin Institute of Technology, Harbin 150001, China）
出处
《智能计算机与应用》
2011年第2X期14-19,共6页
Intelligent Computer and Applications
