Due to nonlinearity, system temperature drift and other uncertainties of PZT (PbZrTiO3), there exists uncertainty for the position of AFM (Atomic Force Microscopy) tip in the task space. It seriously affects the observation and operation efficiency of AFM tip. Thus, to reduce the uncertainty of tip position, and achieve precise positioning becomes an urgent issue to be solved. Firstly, this work represents the uncertainties of tip position in the task space by using the probability distribution. Then, tip motion model is established, and local scan based observation model is combined to estimate optimal tip position by using the Kalman filter. In addition, a model parameter calibration scheme is designed to implement the proposed method. The validity and feasibility of the algorithm are verified by the simulation and experimental results. AFM tip precise positioning can be realized in the task space, and the efficiency of nanomanipulation can be improved.