Design Of Intelligent Monitoring Platform Based On Deep Learning

With the continuous development of China's social economy,the social demand for security products is increasing,which drives the rapid development of the security industry.As an important part of security products,video monitoring system is changing towards to the direction of digitalization,networking and intelligence.However,the development of traditional monitoring system is limited to the upgrade of transmission mode and the improvement of clarity.The post-processes of monitoring videos are still at the level of saving videos,besides,they do not make the best of the effective information in monitoring video.Moreover,with the rise of deep learning algorithms,they become efficient and convenient to extract effective information from images.At the same time,with the development of intelligent office and Internet of Things(IoT),in order to facilitate the management of employees,many companies have equipped their staffs with positional information devices.How to make full use of all kinds of monitoring information to build an intelligent monitoring platform is an important problem in theory and practice.Therefore,an intelligent monitoring platform is designed in this thesis.The purpose is to combine the information extracted from image and personnel positional information on the basis of making full use of the image information in the monitoring video.Then,we establish the video index with the fusion result so as to reduce the human resources consumed in the monitoring video.The main works of this thesis are as summary follows:(1)In order to make full use of the image information in the monitoring videos,this thesis investigates the commonly used pedestrian detection and tracking algorithms,and proposes targeted solutions for the problems of high camera installation position,small target and others in the monitoring video.Through the retraining of detection and tracking algorithms,the performances of the above two algorithms are improved and the performance requirements of intelligent monitoring system are satisfied.(2)In order to integrate personnel position information with image information,this thesis proposes an algorithm to convert geographic position coordinate into image pixel coordinate based on the principle of camera imaging,and solves the distortion problem of surveillance cameras by calibrating cameras.(3)Through the research on pedestrian detection and tracking algorithms and coordinate transformation algorithm of position and pixel.This thesis proposes a joint algorithm that can integrate the above three algorithms.Then,we conduct feasibility analysis and performance test on this algorithm.(4)By using the results of the joint algorithm,this thesis designs a monitoring index system based on the person's name or time and a monitoring video multi-level storage system.This thesis also builds a display interface for the intelligent monitoring platform,which is convenient for retrieving video processing results and monitoring video in real time.In conclusion,based on the research of visual perception technology,this thesis builds a joint algorithm that can be used in the intelligent monitoring system by integrating the location information of personnel.Then,we apply this algorithm to build an intelligent monitoring platform.A new idea is provided for the design of intelligent monitoring.