The invention discloses a free-form surface detection method which includes the following steps: using three-coordinate measuring equipment to measure and calibrate the structural position parameter Sof a reverse Hartmann test optical path including a to-be-detected reflective spherical surface, a projection screen and a CCD camera; determining surface shape error data W0 of the to-be-detected reflective spherical surface according to the measured and calibrated structural position parameter S; determining the surface shape deviation optimization objective of the to-be-detected reflective spherical surface through Zernike fitting according to the measured and calibrated structural position parameter S and the surface shape error data W0; and determining the deviations of the initial measured and calibrated structural position parameter S according to the determined surface shape deviation optimization objective, and determining the actual surface shape error Delta(W) of the to-be-detected reflective spherical surface according to the deviations. By using the method, the surface shape measurement error of the free-form surface is reduced.