基于深度学习的视频监控预警系统

Video Monitoring and Early Warning System Based on Deep Learning

视频监控被广泛应用于社会治安、道路交通、智能家居等场景,但目前部署的视频监控普遍智能化程度还不够高,实时监管、影像回放、发出警报等系列动作一般由专人操作,不仅耗费大量人力、物力和财力,而且极有可能出现迟报、漏报、误报等情况。为解决上述问题,现提出一套基于深度学习的视频监控预警系统,将不同类型的监控视频区分为盗窃抢劫、打架斗殴、交通事故、非法聚集、应急救援五种具体场景,利用预先训练好的深度学习网络模型对监控视频进行特征提取、分类预测,形成预警信息后推送至监控视频调度中心,为实现监控视频自动处理提供一种有效可行的方法。

Video monitoring is widely used in social security,road traffic,smart home and other scenes.However,the current installation and deployment of video monitoring is not intelligent enough.A series of actions such as real-time supervision,video playback and alarm sending are generally operated by special personnel,which not only consumes a lot of human,material and financial resources,but also is very likely to have delayed reporting,missed reporting,false reporting and so on.In order to solve the above problems,a set of video monitoring and early warning system based on deep learning is proposed.Different types of monitoring videos are divided into five specific scenarios:theft and robbery,fighting,traffic accident,illegal gathering and emergency rescue.The pre-trained deep learning network model is used to extract features,classify and predict the monitoring videos,form early warning information,and then push it to the monitoring video dispatching center,it provides an effective and feasible method to realize the automatic processing of monitoring video.