Research And Implementation Of Object Trackingoriented Spatio-temporal Data Processing Technology

With the rapid development of the Internet of Things technology and the rapid deployment of 5G base stations,more and more artificial intelligence applications based on deep learning are being made,and medical fields such as medical imaging,urban security,and autonomous driving based on spatio-temporal data have become the research hot spot in the Internet of Things(Io T)environment.As the number of Io T nodes increases and the network deployment among nodes becomes more and more complex,the video data and picture data collected by the Io T system also increase dramatically at the same time,eventually making the processing speed and feedback speed of the Io T system slower and slower.In order to achieve fast target tracking and and accurate feedback,efficient Io T spatiotemporal data processing methods become particularly important.Therefore,how to reasonably classify and store these collected Io T spatiotemporal data,and correct the transient abnormal data obtained by the Io T become very important.So the Io T query system can search the spatio-temporal data transmitted by the Io T nodes faster information.Based on the spatio-temporal data of the Internet of Things,this thesis will study the correction method for abnormal data collected by terminal nodes and simultaneously implement a classification algorithm for spatio-temporal data.Also,a targetoriented fast search algorithm is implemented.The main research results of this thesis are as follows:(1)Design of Spatio-temporal Data Processing Algorithms for the Internet of ThingsAiming at the problems of abnormal data,large amount of data,and real-time nature of Io T nodes,an EPLSN algorithm based on long-short memory network is designed to solve the problem of classification of transient abnormal data and spatio-temporal data.At the same time,a temporal database is used for the temporary storage of spatio-temporal data,and an Io T search architecture based on spatio-temporal data is designed to be suitable for the real-time search of the Internet of Things and speed up the real-time search.(2)Design of target tracking algorithm based on spatio-temporal dataAiming at the overly complicated target tracking algorithm model and the problems of displacement,occlusion,and illumination in target tracking,a BLMDNet based on the MDNet model are proposed.The model expands the training samples,and improves the focus of target tracking by modifying and expanding the number of convolution kernels.The experimental results show that the accuracy of the target tracking of the proposed BLMDNet model is higher than that of the CNN-SVM model by approximately 0.2.(3)Design and implementation of Io T search system for target trackingBased on the Io T spatio-temporal data,an Io T search system for target tracking is designed,including requirements design,system function design,Io T spatio-temporal data platform construction,architecture design,etc.This thesis briefly introduces the hardware required for the prototype system of the Internet of Things,and implements the target tracking-oriented Io T search system.The system can realize target tracking and real-time monitoring of the positions of sensor nodes and cameras,and can query the temperature,humidity,and gas concentration of the target position.The feasibility of the EPLSN algorithm and the BLMDNet algorithm is also verified.