It is required to carry out on-line quality detection on the processing components in advanced manufacturing. The fringe projection three-dimensional measurement technology is adopted because it meets the requirements of on-line and non-contact. However, due to the large difference of surface reflectivity of metal materials, the phenomena of overexposure and underexposure will occur in the captured structured images and result in the missing or mistaken of the final measuring data. Aiming at this problem, the point cloud data of the estimated shape of the measured object in the camera coordinate system is obtained by iterative nearest point (ICP) point cloud registration based on the incomplete point cloud data measured directly and the designed data of the components obtained by computer-aided design (CAD). Taking the data as the estimated shape, combined with the system calibration parameters, the image intensity correspondence of the same point between projection and imaging is established. The projection gray range and the minimum projection grayscale are calculated by using the intensity imaging response curve of each point on the measuring element by camera. Finally, the three-dimensional shape measurement of high reflective elements can be realized by using a projection grating with non-equivalent coefficient. The experimental results show that the proposed method can completely measure three-dimensional shape of high reflective elements without changing the structural parameters of the measurement system nor increasing the structural complexity of the measurement system.