Intention Discrimination Of Moving Targets In Battlefield Based On Attention Mechanism

With the rapid development of computer vision technology,the scope and depth of artificial intelligence applications are also expanding.Computer vision technology’s utilization in the military for the purpose of intelligent operations has become a paramount component of military construction in numerous nations.Discriminating the intention of moving targets on the battlefield is an important condition in the field of battlefield situation development.It is of guiding significance for us to formulate combat plans to quickly and accurately obtain the enemy’s combat information on the battlefield,understand the opponent’s combat intention,and realize classification recognition and intention judgment of moving targets.At present,many scholars have conducted research in this direction,but there are still some problems to be solved.The traditional intention discrimination model often relies on expert experience,and is easily affected by human emotional factors,which leads to a certain deviation between the results and the actual situation.Aiming at the shortcomings of existing methods,this article proposes an AMCPSO-Bi GRU battlefield tank target discrimination model based on historical trajectory prediction.It combines complex battlefield information features with deep learning technology to achieve accurate identification of battlefield mobile tank target intentions.The main work of this article includes the following:1)In order to improve the speed of battlefield target intention recognition,this paper takes the gated recurrent unit network(Gated Recurrent Unit,GRU)as the basic network unit in the intention discrimination model.A comparison between GRU and the Long Short-Term Memory Network(Long Short-Term Memory,LSTM)reveals that GRU’s performance is comparable,yet its structure is simpler and its training speed is faster.Therefore,the use of GRU can effectively reduce the network training complexity of battlefield target intention recognition.And this article adds a two-way propagation mechanism to the GRU network.Compared with the traditional GRU,the two-way gated recurrent unit(Bidirectional Recurrent Neural Network,Bi GRU)can link past and future information,send it to the network for training,improve the network’s learning ability,and add a zoneout layer between the network layers to prevent the long-term memory stored in the GRU unit from being lost.2)This article introduces the attention mechanism to the trajectory prediction and intention recognition processes in order to handle the intricacy of the battlefield environment and the variety of data.This mechanism can more effectively capture the long-term dependencies,and can ignore those factors that are not related to the prediction results.At the same time,the model will be adjusted according to the weight of features,so that it can focus more on those factors that contribute to the prediction,and obtain more accurate results.In this paper,an improved particle swarm optimization algorithm is introduced into the intention discrimination model to solve the problem of long training times caused by the cumbersome process of parameter optimization.3)A framework for the AM-CPSO-Bi GRU battlefield moving target discrimination model based on historical trajectory prediction is built.On the data set obtained by the combat simulation platform,the experimental simulation analysis is carried out using trajectory prediction,different improved algorithms of LSTM,and the fusion of various battlefield influencing factors.Finally,the AM-CPSO-Bi GRU model is compared with the other four classical network models in the simulation data set.The simulation results show that the accuracy of AM-CPSO-Bi GRU discriminant model is 97.7% compared with other methods.