Font Size: a A A

Software Development Of Visual Intercom System With Face Recognition For Police

Posted on:2020-08-16Degree:MasterType:Thesis
Country:ChinaCandidate:H L LuoFull Text:PDF
GTID:2416330572988044Subject:Electronic information technology and instrumentation
Abstract/Summary:PDF Full Text Request
In recent years,prisons and detention centers are advancing the informationization of security work but there are still two prominent problems.Firstly,the traditional electronic security equipments such as walkie-talkies have many problems.Secondly,the lack of identification leads to unauthorized substitute and off-duty issues.There is a risk of information leakage when the prisons use the visual intercom system,which is developed by foreign companies.The development of a visual intercom system based on domestic chips that meets the needs of constabulary information security and identity verification can solve the two problems mentioned above and has high engineering application valueThe system is designed under Android platform.The RK3399,which is produced by the Rockchips,is used as hardware support platform.The system includes multiple subsystems and functional modules.Visual communication subsystem provides real-time multimedia communication based on SIP protocol and multimedia codec technology to improve multi-point linkage efficiency.The information review subsystem implements real-time storage of call records and remote access and download based on audio-video mixed-stream technology,providing supervision of security equipments.The maintenance subsystem issues alarm events,which facilitates the supervisor to timely master unauthorized substitute and off-duty issues.In view of the needs of identity recognition and the performance constraints of RK3399,the face recognition subsystem designed in this thesis includes Viola-Jones-based face detection module and CNN-based face recognition module,which uses the improved Softmax classifier to train the CNN.According to the test,system's visual communication video frame interval is 59.1ms,the frame's mixed-flow time of call record is 3.04ms,the frame rate of face detection is up to 30FPS,the entire face recognition process takes about 104ms in the shortest time and the accuracy of parameter model is 97.15%.
Keywords/Search Tags:Face Recognition, Visual Intercom, Convolution Neural Network, RK3399
PDF Full Text Request
Related items