Abstract: CR-InSAR is to lay a number of artificial scatterers in the study area to overcome the spatial and temporal decorrelation effect on D-InSAR and to detect high precision deformation of complex terrain area. However, when the deformation between two CR points is more than half radar wavelength, phase unwrapping is becoming extremely difficult. At present, common phase unwrapping methods used for CR-InSAR are Lambda algorithm, minimum cost flow algorithm, et al, but these algorithms all require successfully unwrapping.In the paper, we adopt an improved nonlinear artificial corner reflectors InSAR algorithm based on phase coherence. This method can not only avoid the phase unwrapping error effectively, but also has the strong capability to detect the large scale deformation gradient. In the separation of nonlinear deformation, according to the different characteristics of atmospheric phase in special and temporal domain, taking atmospheric phase and noise phase together as random error, applying the least square method to calculate nonlinear deformation parameters to reduce the loss of nonlinear deformation as far as possible. Taking Jiaju landslide located in Danba County of Sichuan Province as an example, we obtain the deformation of the corner reflectors placed in Jiaju landslide. By comparing the result with that of the SBAS InSAR algorithm and the unwrapping phase based CR-InSAR algorithm, it proves that the improved method can obtain the nonlinear time series result accurately and also reveals that Jiaju landslide has obvious deformation characteristics reasonably.