Research On Video Processing Technology Of Network Conference Based On Panoramic Vision

With the rapid development of information technology and computer network technology,network video conferencing has become one of the mainstream communication methods.The limitation and intelligence of field of view are the main problems that must be solved in its development.In order to enlarge the conference field of view and identify the conference personnel by face recognition technology,the video of the network conference is collected by the catadioptric panoramic vision system.The main work includes:1.The project design of network video conference system based on panoramic vision.Firstly,this paper analyzes and selects the panoramic vision system,and on the basis of choosing the reflected panoramic vision system and analyzing the network structure of network video conference system,The design of network video conference system based on panoramic vision is carried out,and the components of the system and the functions of each part are analyzed.This paper designs the video capture system and analyzes the technology needed to realize the identification of the meeting personnel.2.The detection algorithm of conference personnel is studied.In view of the complexity of the actual video conference environment,the preprocessing of cylindrical panoramic video is carried out,including gray level transformation and histogram equalization of each frame of panoramic image,and the human eye localization method is adopted.The problem of plane rotation in face image is solved.In order to obtain the location of the conference personnel in the video conference,the Adaboost algorithm is used to detect the human face of the conference video.Aiming at the problem that the high resolution of the panoramic image makes the computation of feature extraction large,an algorithm is proposed to reduce the detection area.Finally,the results show that the algorithm is robust to the influence factors such as light and occlusion in video conference.3.Based on the analysis of PCA(Principle Component Analysis(Principal component Analysis)and LBP(Local Binary pattern(Local binary pattern)feature extraction algorithm,the principle and advantages and disadvantages of the algorithm are analyzed.This paper presents a robust algorithm for identification of conference personnel based on the fusion of PCA and LBP,which is robust to the changes of illumination and meeting environment.The comparison experiment shows that the proposed algorithm has the following characteristics: 1.Better robustness and higher recognition rate.4.According to a series of strategies proposed in this paper,a network video conference application software based on panoramic vision is developed,which can obtain 360 °conference scene information and identify the identity of the meeting site personnel in real time.According to the different participants,the network conference application experiment based on panoramic vision is carried out.The real-time detection and recognition success rate of the conference personnel are fully verified.The results show that the proposed face recognition algorithm is effective.It has better recognition effect in the actual meeting environment.