群体异常监测系统中特征点的提取与优化研究

Research on Extraction and Optimization of Feature Points in Swarm Anomaly Monitoring System

针对公共场合安全监控等方面的应用需求,论文研究并设计实现了一个人群突发异常行为的监测系统。目前应用的大多数系统主要是对运动对象进行监测和跟踪。为提高监测的精确度,提升监测效率,在监测过程中通过对特征点进行判断处理会得到更好的效果。所以在样本中选取合适、准确的特征点是关键。论文主要从某一范围内高度唯一性和处于运动前景区域的并具有良好运动属性的特征点两方面进行判断选取。根据判定条件,依据Harris特征点提取算法对特征点进行初步提取,再通过混合高斯背景建模算法对特征点进行进一步的筛选。

In view of the application requirements of safety monitoring in public places,this paper studies and designs a monitoring system for sudden abnormal behavior of people.Most of the systems used at present mainly monitor and track moving objects.In order to improve the accuracy and efficiency of monitoring,better results can be obtained by judging and processing feature points in the process of monitoring.Therefore,selecting appropriate and accurate feature points in the sample is the key.The paper mainly selects the feature points which are highly unique in a certain range and which are in the motion foreground region and have good motion attributes.According to the decision conditions,the feature points are extracted preliminarily according to the Harris feature point extraction algorithm,and then the feature points are further screened by the mixed Gaussian background modeling algorithm.