Abstract: In large numerical aperture (NA) spherical surface shape detection, the defocusing error introduced by the measured sphere deviating from the ideal confocal position will seriously affect the detection accuracy. In order to accurately extract and effectively correct defocus errors and improve the efficiency of adjusting the final inspection position of components, this paper designs a defocuing error correction method based on the error extraction algorithm assisted installation and adjustment. The defocusing error extraction algorithm is used to quantitatively extract the defocusing amount and determine the direction of adjustment, so that the large NA spherical surface can be adjusted to zero fringe position quickly and accurately, and the defocusing error can be greatly reduced. After auxiliary installation and adjustment, Zernike wavefront fitting algorithm is used to eliminate the defocusing term and the spherical difference term introduced by the large NA sphere for fine adjustment. Using this method to correct the large NA spherical surface, the extraction error of defocus is less than 3%. The results show that this method increases the range of defocus error extraction and compensation for components, and can extend the upper limit of defocus to ±25 μm, and ultimately achieve the sub nanometer correction accuracy.