The detection method of combining laser and eddy current is often used in industrial specimen coating thickness measurement. A correction method based on adaptive multiple regression analysis is proposed to solve the problems of correction difficulty and unsuitable correction when measuring variable curvature workpieces. Firstly, the effectiveness and complexity of eddy current measurement correction are analyzed by traditional Gaussian function and polynomial fitting method. A second-order polynomial fitting equation with adaptive parameters is introduced to characterize polynomial parameters and radius of curvature. Secondly, an adaptive surface fitting correction model is established to further simplify the complexity by extending the measurement distance and workpiece curvature. Finally, the introduction of a linear regression model simplifies the related parameter variables of the adaptive surface and improves detection efficiency. It is proved by experiments that the adaptive multiple regression correction method proposed in this paper improves the applicability of curved workpieces and provides a new idea for the nondestructive testing of non-planar products in engineering practice.