A method of calibrating static aberrations in non-common optical path of adaptive optics (AO) system by using phase diversity (PD) techniques is proposed. PD techniques are used to estimate the distorted wavefront phase and recover the object imaging by exploiting the collection of two or more short-exposure optical images, one of which is formed by focusing and others by defocusing. In order to compensate the non-common path error, PD algorithm is used to detect the static aberrations of the imaging optical path online when the AO system is working in close-loop state. Then the result is transformed to the deformable mirror′s initial surface. Experimental results indicate that the image quality is improved obviously after calibration. The value of full-width at half-maximum within the object is reduced by about 14% and the remainder error of the system is reduced by about 72%. The remainder error detected in the imaging optical path is consistent with that of the AO close-loop path, and the ability of the PD techniques is verified in optical detection. The proposed method does not need to change the AO optical path and can accurately figure out the system static aberrations in high signal-to-noise ratio (SNR) conditions, so it can be used as a good mean to detect optical path error in the large aperture telescope.