Research On Terminal Trajectory Analysis Technologies Based On Artificial Intelligence

In recent years,with the vigorous development of wireless sensor networks,the collection and transmission technologies of target data have become more and more mature.Benefiting from this phenomenon,terminal trajectory analysis related technologies can explore the movement rules and behavior patterns of target objects by extracting and mining the spatio-temporal dimension characteristics of massive trajectories,and can be used in civil fields such as smart transportation and urban planning,as well as in military fields such as combat group movement reconnaissance and flight trajectory monitoring.It has important application value in the field.However,the diversity of terminals and the huge size of trajectory data pose great challenges to trajectory analysis.This thesis studies the terminal trajectory analysis technology based on artificial intelligence for the terminal trajectory data collected by wireless sensors.The specific work and contributions are as follows:Firstly,this thesis studies the key technologies of trajectory analysis such as trajectory data equalization,trajectory terminal classification and trajectory similarity analysis,and introduces the working principle and advantages and disadvantages of commonly used data equalization algorithms,the mathematical model and implementation method of time series classification based on deep learning,and The main principles and application scenarios of trajectory compression,trajectory clustering and trajectory similarity measurement algorithms.Secondly,in order to extract the terminal of the target trajectory,this thesis designs a terminal classification scheme based on deep learning.Among them,aiming at the imbalanced problem of actual collected data,a random interval sampling method is proposed to balance the trajectory data volume of different types of terminals.Aiming at the problem of uneven sampling time interval of actual trajectory data,this thesis proposes an improved low-complexity classification network model based on the Time-LSTM network structure.The experimental results show that the data balancing scheme proposed in this thesis can effectively balance the data.When compared with the traditional algorithm,the improved classifier has better real-time performance under the premise of ensuring the classification accuracy.Finally,this thesis proposes a low-complexity trajectory similarity analysis scheme with three stages,including data preprocessing,trajectory clustering and similarity judgment.In the data preprocessing stage,the redundancy is reduced based on the scalable window trajectory compression algorithm,so as to initially reduce the calculation amount of subsequent analysis.In the trajectory clustering stage,a trajectory similarity measurement method based on trajectory segments is proposed to reduce the complexity of trajectory clustering and subsequent trajectory similarity judgment.In the similarity judgment stage,a two-stage similarity judgment strategy is proposed to further reduce the computational complexity.The experimental results show that compared with the traditional scheme,the scheme proposed in this thesis can improve the accuracy of trajectory similarity analysis while reducing the calculation time.In summary,the artificial intelligence-based terminal trajectory analysis scheme proposed in this thesis is designed with full consideration of the actual application scenario requirements and the imbalanced characteristics of the actual collected data,and is verified experimentally based on the actual data collected by practical wireless sensors.The results show the high accuracy and low complexity of the scheme,which has promising practical application value.