The PHD/CPHD filter with the adaptive target birth intensity driven by measurements is proposed. The result that the newborn targets are not always detected can solve the problem of normalized unbalance. The adaptive target birth intensity can be designed based on measurement-driven and the estimated state and number of persistent targets, and the newborn targets are propagated separately by using the PHD/CPHD filter. The SMC implementation of the improved PHD/CPHD filter is described. The numerical simulation results show that the improved algorithms can efficiently and instantaneously estimate the number of targets and their states, and have great application prospection.