Research On Retrieval Technology Of Visual Data Based On Network Element Virtualization

Visual data retrieval techniques are an important research direction in the field of computer vision,with important theoretical significance and technical challenges.Therefore,it is important to study visual data retrieval techniques to improve the efficiency of image and video data acquisition,analysis and application.The existing research on visual data retrieval techniques usually suffers from low retrieval efficiency,heavy computational burden of the retrieval process and difficulty in extracting highlevel semantic information.To address the above problems,the main research of this paper is as follows.(1)Virtualized Network Element-based Visual Data Retrieval Framework(VNEVDRF)is proposed.Combining the high utilization of device and flexibility of NFV technology and the advantages of automation,flexible management,and high fault tolerance of Kubernetes,the proposed VNE-VDRF can further improve the management efficiency of visual data retrieval service providers and achieve more flexible and efficient application deployment,management,and operation.This virtualization and containerization approach can improve the performance,reliability,and scalability of the application.(2)A feature classification-based image retrieval method is proposed.Through the design of the retrieval process,the data security of image retrieval is improved from the process,and the progress of reducing the computation and improving the retrieval efficiency is achieved,which plays a key role in further migrating the client-side computation overhead to the cloud and improving the efficiency of image data security retrieval.(3)A scenario information detection algorithm is designed.By establishing a mapping between different entities under the same time and space,it achieves the task of detecting situational information in video data by obtaining the corresponding highlevel situational information through the low-level pixel information of video data.The spatio-temporal correlation of video entities is used as the basis for scenario information detection,and the concept of scenario features is proposed to guarantee the logical accuracy of the algorithm.(4)A video retrieval model based on situational information is established.On the basis of situational information detection,it further accomplishes the tasks of cutting the video based on situational information changes,extracting situational frames based on situational information,and realizing the retrieval of video data based on situational frames.At the same time,the retrieval process is adjusted accordingly with reference to the image retrieval method,which further improves the retrieval efficiency and security.