基于智能视频监控的人流量统计系统设计

发布时间：2018-06-07 01:17

  本文选题：智能视频监控 + 人流量统计　； 参考：《合肥工业大学》2016年硕士论文

【摘要】：人流量统计问题是当前智能监控领域的前沿性课题,涉及到诸多学科和众多复杂的问题。在智能视频监控下对人流量进行统计不仅能应用在商场超市等商业领域,还能在安防、交通等方面发挥重要作用。基于智能视频监控的人流量统计系统是利用智能视频监控技术对监控区域的人流量进行统计,其核心技术包括运动目标检测、人体目标识别和跟踪。本文针对开放环境下的人流量统计需求,设计了基于智能视频监控的人流量统计系统,并对其中的关键技术进行了分析和研究,主要的工作如下：(1)本文在分析了多种运动目标检测算法后,通过对当前主流背景建模算法的评估和比较,确定了基于ViBe的运动目标检测算法。针对传统ViBe算法的不足之处,提出了一种改进的算法,通过对前景像素点做鬼影判定进行二次分类,有效地消除了鬼影并抑制了闪烁点像素噪声。(2)在人体目标识别和跟踪方面,以HOG特征作为行人描述特征,结合改进的ViBe算法得到可能运动区域的HOG特征值,然后通过事先训练好的SVM分类器进行分类识别。本文采用Cam-shift跟踪算法实现对人体目标的跟踪,针对目标被遮挡时出现的无法匹配的现象,结合Kalman滤波器对目标运动轨迹进行预测。(3)基于上述关键技术的研究,本文最终完成了系统软、硬件平台的设计,并开发出客户端界面。系统经过调试和实验,实现了对人流量的准确统计,具有较好的鲁棒性。
[Abstract]:The problem of human flow statistics is a leading subject in the field of intelligent monitoring, which involves many disciplines and complex problems. Under intelligent video surveillance, the statistics of human flow can not only be used in commercial fields such as shopping malls and supermarkets, but also play an important role in security, traffic and so on. The human flow statistics system based on intelligent video surveillance is to use intelligent video surveillance technology to calculate the flow of people in the monitored area. The core technologies of the system include moving target detection, human body target recognition and tracking. In order to meet the demand of human flow statistics in open environment, this paper designs a human flow statistics system based on intelligent video surveillance, and analyzes and studies the key technologies. The main work is as follows: (1) after analyzing a variety of moving target detection algorithms, through the evaluation and comparison of the current mainstream background modeling algorithms, we determine the moving target detection algorithm based on ViBe. In view of the shortcomings of the traditional ViBe algorithm, an improved algorithm is proposed, which can effectively eliminate the ghost shadow and suppress the pixel noise of the flashing point in the aspect of human body target recognition and tracking by making a second classification of the foreground pixels. The HOG feature is used as the pedestrian description feature, and the HOG eigenvalue of the possible moving region is obtained by combining the improved ViBe algorithm, and then it is recognized by the pre-trained SVM classifier. In this paper, Cam-shift tracking algorithm is used to track human body target. Aiming at the unmatched phenomenon when the target is occluded, combined with the Kalman filter to predict the moving trajectory of the target, the research is based on the above key technology. Finally, the software and hardware platform of the system is designed, and the client interface is developed. After debugging and experiment, the system realizes the accurate statistics of the flow of people, and has good robustness.
【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TN948.6
，

本文编号：1989020