The Research On Intelligent Network Monitoring System Structure And Intelligent Analysis Algorithm

With the improvement of people's safety requirements for living places and the rapid development of community economy,monitoring system has been widely used in public security monitoring,industrial site monitoring,residential community monitoring,traffic condition monitoring and other scenes and plays an important role in crime prediction,traffic control,accident prevention and detection,old and weak sickness monitoring,environmental monitoring and so on.The traditional monitoring system only provides basic functions such as real-time capture,storage,and playback of video.It often requires staff to find interesting targets and respond accordingly.This kind of traditional monitoring system that uses manual methods consumes manpower.When the staff is fatigued or neglected,they often cannot deal with the abnormalities in time,thus there is a certain safety risk.Since the end of the 20 th century,with the rapid development of machine learning,big data,and deep learning,the monitoring system has begun to enter the “intelligent” era.The so-called intelligence,that is,the use of computers to replace the human brain for reasonable data processing,the use of camera to replace human eyes by which we can actively seek out and identify the target of interest.Intelligent monitoring system can assist staff to complete a large number of tasks,reduce manpower requirements,reduce monitoring costs,reduce security risks,but also greatly improve the efficiency of monitoring,with extensive research significance and application prospects.In the research of intelligent network monitoring system,its structure and intelligent analysis algorithm are important research contents.In order to design a reasonable and efficient system structure,find the most appropriate intelligent analysis algorithm,the main content of this article includes: 1)Build a monitoring system network structure.2)The software architecture of intelligent monitoring system is studied,and the functions of client and server are roughly divided.3)In order to carry out target detection,a variety of target detection methods were studied,implemented and compared.The Vi BE,a moving object detection method based on background modeling,not only has better performance,but also has better robustness.The global rigid template method based on the target modeling method is relatively fast in rate,and the training is relatively simple and can meet the general requirements.When the hardware conditions permit,the DMP method and the R-CNN series method and the YOLO series method that based on deep learning can obtain high accuracy and robust performance.4)In order to track the target of interest,a variety of target tracking methods were studied,implemented and compared.The generated ASMS method and the discriminative correlation filter method such as DCF,CN,and Staple have better performance and speeds in the single target tracking methods.In order to ensure realtime performance,multi-target tracking adopts a method based on Vi BE and Kalman filtering.In the case where the target trajectory is relatively independent,better results can be obtained.5)Face recognition technology is researched.Aiming at the great influence of illumination on face recognition,an illumination-invariant face recognition algorithm is proposed.After testing on the extended Yale B dataset,it is found that the algorithm achieves more than 98% of face recognition accuracy and has a certain degree of rotation invariance in the case of a few training samples and very poor lighting conditions,and it also has relatively good robustness to face expression.