To achieve accurate measurement of the surface shape of a cylindrical lens, research has been conducted on the measurement errors caused by reflectance. The working principle of the spectral confocal displacement sensing system is first introduced, and the spectral response of the system is derived in the spatial coordinate system using the structural parameters of the cylindrical lens. Then, theoretical research and simulation analysis are conducted on the spectral peak drift caused by the reflectance at each point and the resulting measurement errors in the surface shape. To correct the influence of lens surface reflectance on the measurement results, an error correction algorithm is proposed using S-G filtering and Gaussian fitting to achieve filtering and extraction of the peak wavelength of the spectral signal. Finally, by building an experimental platform, the measurement of the cylindrical lens surface shape based on reflectance correction is completed. The experimental results show that the mean absolute error and root mean square error of the measurement results before correction are 9.18 μm and 9.79 μm, respectively. After correction, the mean absolute error and root mean square error of the surface shape are 0.71 μm and 0.75 μm, respectively, which improves the measurement accuracy of the surface shape and validates the correctness of the theoretical analysis and the effectiveness of the proposed reflectance correction algorithm.