A Study And Implementation On Target Recognition Based Embedded Platform Surveillance Video Storage System

With the in-depth development of artificial intelligence and computer hardware technology,water surface embedded intelligent video surveillance terminals with target recognition and storage functions have come out.The terminal can be deployed to designated sea areas for video surveillance,recognizes the target appearing in the surveillance video in real time,and the surveillance video can be stored locally for the control center to view.The embedded video surveillance terminal usually adopts the method of cyclic coverage and storage of all surveillance videos,which has the problems of slow search speed,high waste of storage space and network bandwidth,and short retention time of important surveillance videos.This paper designs and implements a key-value storage system based on the water surface embedded intelligent video surveillance terminal to classify and store surveillance video to solve the above problems.First of all,this paper proposes a surveillance video storage strategy based on intelligent video surveillance terminals: the surveillance video is recognized locally in the terminal in real-time,the valid videos containing preset recognition targets are stored,meaningless video clips are discarded,and the user is allowed to choose whether save a certain volume of common video as a supplement.This strategy can effectively reduce the waste of storage space and network bandwidth,extend the retention time of important surveillance video clips.Secondly,according to the storage requirements of valid videos and the surveillance video storage strategy mentioned above,this paper designs and implements a kind of key-value storage system: VideoKV,which is suitable for surveillance video storage and friendly to embedded devices,in order to efficiently store valid videos and meet the requirements of users to quickly query valid videos according to multiple attributes.VideoKV adopts optimization measures such as key-value separation storage,storing index structure in memory and data consistency only,which can effectively solve the problems of write amplification and low query efficiency in mainstream keyvalue storage systems when storing large-size surveillance video clips.Finally,VideoKV was deployed to the Hi3559AV100 embedded platform for functional testing,and the performance comparison test of VideoKV was performed by transplanting the mainstream key-value storage systems Level DB and Wired Tiger as the control group.Experimental results show that the functional interface of VideoKV runs correctly;In the range query workload,the performance of VideoKV is at least 3.55 times higher than that of the control group,and in the mixed workload of writing and range query,the performance of VideoKV is at least 4.47 times higher than that of the control group.