The Design And Implementation Of Key Modules On A Smart Home Monitoring System

With the development of modern technologies,crimes are becoming more intelligent and concealed.Security monitoring has gradually become an essential part of people’s dailylife.Therefore,this paper develops a multifunctional smart home monitoring system.The system not only supports cloud storage and remote access of home monitoring pictures,but can also captures video pictures in real time for face detection.By comparing the information in the database,it can judge the intrusion of outsiders and push alarm information in time.The system also supports online updating of the model library and optimizes the face recognition capabilities of embedded devices.This system can be used in scenarios with high requirements for face recognition and security,such as smart cat eyes,smart home cameras,and smart access control.Our work is as follows:(1)We analyze the needs of smart home monitoring system in detail and develop the software and hardware of this system based on Android embedded platform.We further develop the video monitoring module based on Baidu LSS cloud service.In the video surveillance module,video collection,transcoding,processing,and uploading are implemented.(2)A cloud storage system based on the Hadoop environment has been implemented.This system not only implements the login function of different accounts but also supports download and deletion of video files and can avoid data loss.(3)The mainstream mature face detection and recognition algorithms are studied and their performance is tested and compared.Finally,we choose the proper algorithms to implement functions such as stranger face recognition.(4)A set of model incremental update schemes are designed and implemented on the server side,and incremental patch packages are generated through the server’s difference comparison algorithm.This solution can effectively reduce the flow of the update process,reduce the failure rate and update time of the upgrade,so as to achieve an efficient upgrade of the incremental model in scenarios with poor network environments.(5)We complete testing and analysis of each part of the system.The test results show that our system has achieved satisfactory results in response speed and recognition accuracy,and has good practical values.