Research On Target Intention Recognition Method Based On Bayesian Reasoning

Today,science and technology are relatively developed.Science and technology provide many conveniences for human life.At the same time,science and technology are fully integrated into the field of war.The war has changed from the main form of hand to the game between science and technology.Science and technology play two roles of "spear" and "shield" in the war.China has always been taking the maintenance of world peace and the promotion of common development as its duty,not seeking "spear" sharpness,but seeking "shield" firmness.According to the main idea of peaceful development of our country and with the help of science and technology,thesis proposes a target intention recognition method based on Bayesian reasoning,which aims to provide a new method for analyzing the intention of air targets with the method of artificial intelligence,and to strengthen the shield of protecting the country with the strength of science and technology.However,the traditional air target intention recognition method focuses on data processing and application,not enough on the air the role of target behavior rules in intention recognition,such a method gradually does not meet the needs of analyzing air target intention under the condition of information technology.Aiming at this problem,thesis proposes feature-weighted naive Bayesian network based aerial target recognition algorithm and sequence reasoning based dynamic Bayesian aerial target intention recognition algorithm.The main contents are as follows:Naive Bayes algorithm has stable classification efficiency and high classification speed,and is suitable for incremental training.However,in practical applications,the independence of the characteristic conditions it is based on assumes that the conditions are insufficient,so researchers generally expand from the structure.Feature selection,feature weighting,and instance weighting improve it.Aiming at the conditional independent hypothesis problem of Naive Bayes,thesis proposes a feature weight allocation method based on feature data span and density weighted sum.By analyzing the span and density of the overlapping parts of the feature data,the influence degree of different attributes on the classification is obtained,and the feature attributes are weighted based on this.Feature weighting according to this method will greatly reduce the impact of the correlation of feature attributes.Aiming at the problem of air target intent recognition,the process of air battlefield intent planning is analyzed in depth,and the tactical intent of the target is inferred by reversing the state attributes.In this process,a dynamic Bayesian network is constructed according to the state transition of the target,the target's maneuver is identified,and then analysis is performed to obtain the target intention based on the change of the maneuver.Among them,the target maneuvering recognition part has a low time span due to the large time span.In order to solve this problem,thesis improves the probability transition matrix of the dynamic Bayesian network and proposes a state recorded in the duration of the action.The transition probability calculation method analyzes the state of the target at the end of the current action,establishes a state transition model,analyzes the state of the next stage of the target,and obtains the maneuver action of the next stage of the target.This method can improve the accuracy of dynamic Bayesian in target action recognition process,and then improve the accuracy of intent recognition.