Abstract:

Battlefield target clustering is confronted with the problems of the unknown category number and the lack of effective threshold selection methods. To deal with these problems, an improved space partition based target clustering algorithm is proposed. Firstly, by using the friend-or-foe partition and the combat-unitattribute partition, the clustering target number is reduced to lessen the computational burden. Secondly, through improving the space partition, the optimal selection of the dynamic threshold is implemented to solve the clustering problem with the unknown category number. In particular, it can eliminate the computational redundancy and extract the candidate threshold by introducing the partition independence and the probability interval constraint of inverse cumulative χ² distribution. On this basis, selecting the maximum candidate threshold can effectively filter out the process noise and the observation noise, improving the accuracy of clustering. Simulation results show that the proposed algorithm is effective, stable and real-time under the battlefield environment.