Software Development Of Personnel Identification System In Elevator Based On Video Analysis

With the increasing demand for security prevention,video surveillance equipment is now widely installed in elevator cars.However,traditional video surveillance system usually lacks effective analysis methods and fails to achieve good monitoring effects with manual marking methods.Therefore,realizing automatic identification of registered users,strange visitors and blacklists with algorithm can effectively improve the building security and use experience of elevators,and thus has high engineering application value.This thesis develops a person identification system for elevators based on video analysis.The system performs real-time video monitoring in elevator cars and realizes face detection and recognition.The system can perform real-time transmission and decoding on HD video captured in elevator car,and then analyze the video using deep neural network,realizing face detection and recognition.The system can judge the person in the car as a registered user,a strange visitor or a blacklist according to the recognition result,and then execute different control strategies accordingly.In addition,the system also implements the storage,indexing and downloading functions of surveillance video.The system also implements a voice intercom function to realize voice communication between the car and the duty room.Finally,the system integrates relevant protocol modules to enable communication between the system and external client software and the elevator control system.According to the test,the system can realize real-time transmission,decoding and storage operation of 1080p/25fps/H264 format HD video,and achieve good results of face dection and identification.The processing time of decoding one frame of 1080p/H264 format is 3.61ms,while the processing time of face detection and recognition is 38.2ms,which meets the design requirements and has high engineering application value.