A multiple target state extraction algorithm of probability hypothesis density filter based on Dirichlet distribution is proposed. The instability of Dirichlet distribution with negative exponent parameters is applied in estimating multiple target state by maximum likelihood criterion. Dirichlet distribution drives irrelevant components to extinction while the maximum likelihood solution is being searched by expectation maximum algorithm. In order to balance the successful initialization and the reduction of time cost of algorithm

a k-dimensional tree is applied to initialize Dirichlet distribution. Simulation results show that the multiple target state extraction algorithm based on Dirichlet distribution is superior to the existing algorithms for the probability hypothesis density filter in multiple target tracking.