| With the development of mobile Internet and the popularity of intelligent terminals,the market of video conferencing systems has evolved from traditional hardware video conferencing system only,to a situation that hardware video conferencing systems and software video conferencing systems coexisting.The research and implementation of mobile conferencing software becomes a valuable issue.The focal point of this thesis are implementing a video conference which is close to the actual conference scenes on mobile and helping attendees to improve the understanding to the content of conference.In this thesis,a mobile application for video conference is designed and implemented on Android basing on session establishment and audio and video processing of Linphone.After that,the incomplete face detection problem in video conference process is studied.Object detection network is used to research and model the solution of this problem.In the research and implementation process of mobile application for video conference,the requirements are analysed at first,then the functional module division is expounded which includes the basic function module,the video conference module,the interactive whiteboard module,the face detection module and the communication module.Those modules are designed and implemented in this thesis.After that,the communication interface between the mobile application and the server is described.According to the requirements of the conference scenes,a deep incomplete face detection model is designed and implemented based on YOLOv3 to solve the incomplete face detection problem,and the YOLO_mark tool is modified.The face images are marked as training data,and the available incomplete face detection model is trained.Finally,this thesis expounds the testing process of this application,records the test cases and corresponding test results,shows the user interface,summarizes the work of this paper,analyzes the inadequacies,and points out the improvement direction. |
PDF Full Text Request