Aiming at the problem of camera pixel saturation caused by the high reflective area on the metal surface of a relay, we adopt speckle projection profilometry to detect the magnetic circuit drop height of the relay. The main problem is solved through a detailed analysis of characteristics according to different speckle gray values, speckle sizes, and shapes to effectively distinguish the confusion between the speckle reflection spots and background light reflection spots. The enhanced normalized cross-correlation sub-pixel algorithm is used to get a dense matching pair of points of the right and left images, and three-dimensional reconstruction is based on the triangulation parallax principle. The random sampling consensus algorithm and European clustering algorithm are used to segment the point cloud to obtain the core and yoke iron planes and measure the distance between the two planes. We measured two standard blocks with height difference of 1 mm, and the error is within 2% by the proposed method. The correctness is verified through real-time measurement for a magnetic circuit drop height of the relay assembly line.