When the prior noise statistic is unknown and time-varying, the conventional probability hypothesis density(PHD) filter declines in accuracy and loses targets. An adaptive multiple-model PHD(MMPHD) filter is proposed to estimate the states of targets and their noise variances. The unknown and time-varying noise is estimated based on multiple-models methods, and the MMPHD filter is designed to jointly estimate the target states and the statistics of the noise. The sequential Monte Carlo method is used to implement the MMPHD filter. Simulation results show that the proposed filter can accommodate the unknown measurement variances effectively and improve the estimation accuracy.