For defect detection of specular reflection objects, traditional methods are prone to interference from reflections, have weak recognition capability for tiny defects (such as hairline scratches and subtle Pits), and suffer from insufficient detection accuracy due to reliance on single image processing. To address these issues, a defect detection method based on phase measurement deflectometry and mean curvature is proposed. This method uses the gradient information acquired by phase measurement deflectometry (Phase Measurement Deflectometry) as the basis to compute the mean curvature map, and then achieves accurate defect detection through image processing operations such as denoising preprocessing, histogram equalization, and edge detection. The actual measurement results of the phone case show that the mean curvature of the flat area is 9.935×10^-5, while the mean curvatures of pits at the defect locations are -1.843×10^-3 and -2.060×10^-3 respectively, and the mean curvature of cracks is -4.640×10^-3. The absolute values of the mean curvature in the defect areas are significantly bigger. Compared with the traditional white light illumination detection method, the proposed method can completely render the closed outline of pits and the continuous sharp edges of cracks, effectively avoid the interference of pseudo-edges caused by reflections, and significantly improve the defect recognition accuracy. It can effectively identify defects such as pits and cracks on the surface of phone cases, providing a new high-precision defect detection solution for Industrial products.