Reconfigurable intelligent surface (RIS) can manipulate the wireless propagation environ-ment by smartly adjusting the amplitude/phase in a programmable panel, enjoying the improved perfor-mance. The accurate acquisition of the instantaneous channel state information (CSI) in the cascaded RIS chain makes an indispensable contribution to the per-formance gains. However, it is quite challenging to estimate the CSI in a time-variant scenario due to the limited signal processing capability of the passive ele-ments embedded in a RIS pannel. In this work, a chan-nel estimation scheme for the RIS-assisted wireless communication system is proposed, which is demon-strated to perform well in a time-variant scenario. The cascaded RIS channel is modeled as a state-space model based upon the mobility situations. In addi-tion, to fully exploit the time correlation of channel, Kalman filter is employed by taking the prior informa-tion of channels into account. Further, the optimal re-flection coefficients are derived according to the min-imum mean square error (MMSE) criterion. Numeri-cal results show that the proposed methods exhibit su-perior performance if compared with a conventional channel estimation scheme.