Abstract: ‍ ‍Extremely large-scale massive MIMO（Multiple-Input Multiple-Output） systems have great potential in the development of future communication industry. The large size of the base station antenna array will bring different channel environments for the close communication between the user and the antenna array： the near-field effect and spatial non-stationability of wireless channeland， and will bring more challenges to channel estimation. As to the near-field effect， it is necessary to estimate not only the angle but also the distance， this paper proposed to estimate angle and distance separately to reduce the complexity of estimation. As to the non-stationarity， range of the visible subarray of the scatterer needs to be determined. In this paper， the angle of each subarray was estimated separately， the visible region of the scatterer was judged according to the angle and the receiving signals covariance matrix to do the decomposed eigenvalue， and then the distance was estimated according to the angle obtained. Further analysis shows that the angle which is estimated by the subarray near the center and the distance which is estimated by the subarray near the edge is more accurate than other subarrays. On this basis， the scheme of determining the angle and the distance of the extremely large-scale array under different scattering conditions was given. To sum up， to deal with the near-field effect and spatial non-stationability in wireless channel of extremely large-scale massive MIMO systems， this paper estimated the angle and the distance separately in order to reduce the computation complexity， firstly， the angle was estimated by the subarray which is near the center， and thus the visible region of the scatterer could be judged according to the angle and the eigenvalue， then finally the distance was estimated using the subarray near the edge. In the end of this paper， some simulation results show that the channel estimation scheme proposed in this paper has smaller mean square error performance. Besides， we can also find from the simulation graphy that for angle estimation， the closer the subarray is to the center， the more accurate it is and for distance estimation， the closer the edge subarray is， the more accurate it is， which is consistent with the previous analysis.