The spatial uncertainties of atomic force microscope (AFM) tip position hinder the development of the AFM based nano-manipulation. These uncertainties cannot be corrected at nanoscale using the traditional position sensor, which is used in the macro robot localization. As for that problem, we propose that the AFM tip is used as the sensor to detect the landmark in the sample surface via a local-scan based motion. The landmark positions are used to estimate the tip position in the task frame. The local-scan based observation model is built on the foundation of the tip motion model. These model parameters are calibrated using statistical experiments. Simulation and experimental results show that the proposed method can improve the accuracy of the tip position. Then the influence of the tip position accuracy is analyzed by using nano-manipulation results of the experiments. Furtherly, three important factors in AFM tip based nano-manipulation are discussed. The landmark domain as the first factor is analyzed for assurance of the tip accuracy before nano-manipulation. The second one is studying the contact characteristics between the nanoparticle and the substrate by detecting nano-manipulation force. The last one is taking the tip shape into account for effective manipulation through fine tuning the tip offset. Finally, the experimental results illustrate the effectiveness of the proposed method for fabricating the nano-structures and devices.