Anomaly Detection Of Battlefield Targets Activity Based On Clustering

With the rapid development of modern information technology,various high-tech technologies have been widely applied in the military field,making the war more and more information-based.It is also increasingly important for both parties of the war to master and use information.However,the improvement of the army's information level also makes the battlefield data huge.Therefore,how to deal with a large amount of battlefield information,provide commanders with reliable and useful battlefield information,and help commanders to make correct and effective decision-making become the key to gaining battlefield advantages.In order to detect the sudden changes of enemy strategic accurately and timely on the battlefield,it is very effective to analyze the abnormal changes in the battlefield reconnaissance data.At this stage,however,most information systems mainly present battlefield information to commanders through visualization techniques,the main function of which is to display information,and not to analyze the other valuable operational information hidden in the data mining.It also can not make full use of the battlefield information.Therefore,it is necessary to adopt a more effective computer-aided way to detect abnormal changes of enemy target information data.Based on this,the main research content of this paper is to analyze the abnormal changes of the battlefield target activity information,including the dynamic anomaly of the group target,the statistical distribution anomaly in the frequency domain of the group target,and the dynamic anomaly of the single target,providing theoretical techniques for situation prediction.support.The details are as follows:(1)Studying the method of detecting the dynamic anomaly of the target position in the fixed area of the battlefield.Due to the non-uniform distribution of targets on the battlefield,this paper proposes an improved DBSCAN clustering algorithm to compensate for the poor clustering effect of the traditional DBSCAN clustering algorithm on non-uniform density data,and then designs the anomaly detection process and verifies the effectiveness.(2)Analyzing how to perform abnormal analysis on the statistical quantity in the frequency domain of the target of a fixed area of the battlefield.This paper proposes an improved K-means clustering algorithm based on quadratic distance for the research scenario,which effectively solves the problem that the traditional K-means clustering algorithm can not measure the correlation between the eleven elements in the vector of statistical quantities in the frequency domain,then using statistical methods,designed an anomaly detection process and simulated the results.(3)Designing a dynamic anomaly detection method for a single target in a fixed area of the battlefield.In this paper,through the grid data,the K-means clustering algorithm is used to deal with the anomaly,and a set of anomaly detection process is designed.Then the effectiveness is verified by the simulation results.