基于音视频分析的区域安防管控平台

Regional Security Management Platform Based on Audio and Video Analysis

针对广场等区域人流量大、安保要求高的特点,设计并实现了一套基于音视频识别的区域安防管控平台.该平台从区域安防的需求出发,设计了嫌疑人员报警、人群拥挤报警和异常声音报警三大功能,以人脸识别、人群密度估计、异常声音识别三个音视频识别算法作为支撑,实现对区域内与人相关的异常事件的预警.首先介绍了基于深度神经网络的人脸识别算法,然后介绍了一种引入注意力机制的卷积神经网络模型来实现的人群密度估计算法,和一种基于多卷积神经网络模型融合的异常声音识别算法;最后介绍了平台需求与设计过程,主要包括平台建设的需求分析和界面设计.

We design and complete a regional security control platform based on audio and video recognition in order to protect people in the area from dangerous situation. The platform is designed three functions: suspect alarm, crowd crowding alarm and abnormal voice alarm. It uses face recognition, crowd density estimation and abnormal voice recognition as the support to realize the early warning of abnormal events in the region. Firstly, it introduces the face recognition algorithm based on deep neural network, then a convolutional neural network model with attention mechanism is introduced to realize the population density estimation algorithm. Finally, it introduces an abnormal voice recognition algorithm based on multi-convolutional neural network model fusion. At last it introduces the platform requirements and design process, including the platform construction requirement analysis and interface design.