基于多目标识别的智能安防视频监控研究

Research on Intelligent Security Video Surveillance Based on Multi-target Recognition

针对视频监控受到动态目标影响监控识别效果差的问题,提出基于多目标识别的智能安防视频监控研究。充分考虑多类目标情况,将多目标预判成词袋;分割处理图像,计算目标点与超平面的间隔,聚类处理多目标构建高斯模型表征背景,剔除高斯分布中存在的虚假背景,二值化处理目标,分离目标与背景。对差值进行阈值化判断,结合3×3邻域的中值滤波器去除噪声;判定入侵行为,完成智能安防视频监控识别。实验结果表明,该方法能够识别出全部目标,TA最大值为21%,TN最大值为90%,监控识别效果良好。

Aiming at the problem that video surveillance is affected by dynamic targets and the effect of surveillance recognition is poor,an intelligent security video surveillance research based on multi-target recognition is proposed.It fully considers the situation of multiple targets,and pre judges the multiple targets into word bags,segments the image,calculates the distance between the target point and the hyperplane,and clusters the multi-target.It constructs Gaussian model to represent the background and eliminate the false background in the Gaussian distribution,binarization deals with the target and separates the target from the background.It performs threshold judgment on the difference,combined with 3×3 neighborhood median filter to remove noise,determine the intrusion behavior and complete the intelligent security video monitoring and identification.The experimental results show that this method can recognize all targets,with the maximum TA of 21%and the maximum TN of 90%.The monitoring and recognition effect is good.