基于卷积神经网络的智能家庭安防监控系统

Intelligent home security monitoring system based on convolutional neural network

传统家庭安防监控系统接受率低,无法实时通知用户家庭异常,不能满足用户需求。针对这一问题,文中设计一种基于卷积神经网络的智能家庭安防监控系统。该系统采用卷积神经网络算法对摄像头采集到的图像进行检测识别,可对人和动物等多目标进行快速识别区分,避免传统检测方法中使用人体红外感应传感器造成的误判。当系统检测到有人闯入时,单片机控制蜂鸣器报警,且通过GSM模块将识别结果和视频链接以短信方式发送到手机端,使用者可通过点击视频链接及时查看现场情况。通过将卷积神经网络和单片机结合,该系统可完成实时目标检测并发送短信给用户,用户可以实时调用视频查看异常,从而减少用户生命财产受损,该系统成本低,用户接受度高。实验结果表明,该智能安防系统工作稳定可靠,误报率低,可对居家环境智能化实时监控和报警。

The traditional home security monitoring system has the low acceptance rate,cannot inform users of the abnormal situation in real time,and cannot meet users′needs.On this basis,an intelligent home security monitoring system based on convolutional neural network is designed.In the system,the convolutional neural network algorithm is adopted to detect and recognize the images captured by the camera,with which the multiple objects such as humans and animals can be quickly recognized and distinguished to avoid the error judgment caused by human infrared sensor used in the traditional detection method.When the system detects an intruder,the MCU(microprogrammed control unit)can control the Buzzer to give an alarm,and send the recognition result and video link to the mobile phone by SMS(short messaging service)through the GSM module.The user can check the abnormal scene situation by clicking the video link.The real⁃time target detection and shot massage sending to users can be completed by the system in combination of the convolutional neural network with MCU.The users can call the video to view exceptions in real time,so as to reduce their loss of life and property,which has low cost and high user acceptance level.The experimental results show that the intelligent safe⁃guard system can work stably and reliably,make real⁃time alarm and has low false alarm rate.