During the measurement of surface topography on objects with high-reflective surface by numerical grating projection three-dimensional measurement technology, the acquired images may exhibit local overexposure or underexposure, which consequently disrupts the projected grating pattern and ultimately leads to a decline in measurement accuracy.In order to accurately measure the three-dimensional shape of objects with high-reflective surface, an improved automatic multiple exposure image fusion technique was proposed. Firstly, the appropriate measurement grayscale range was determined by fitting the grayscale response curve of the system, which further led to the determination of the exposure time range. Within the exposure time range, the exposure time was gradually increased with certain step size, and the camera synchronously captured and saved images at each exposure time. The number of points within the determined grayscale value range was statistically calculated by traversing the images. The image with the highest number of points within this range was selected as the datum exposure image, and its corresponding exposure time was considered the datum exposure time, while the point in the image with the smallest pixel grayscale value within this range was identified as the datum point. Secondly, the camera response function was fitted based on the transformation of the pixel grayscale value of the datum point with exposure time. According to the camera response function, the surface illumination value range of the measured object can be calculated, and the required exposure time and frequency for the current object can be determined based on the idea that different exposure times correspond to different illumination value ranges. Subsequently, a CCD camera was utilized to capture pure white images and grating pattern images projected onto the target object under various exposure durations. The images obtained from the pure white projection were utilized to generate masks, which were sequentially multiplied with their corresponding grating pattern images. The obtained images were superimposed after linear transformation to ensure that no additional highlights are generated near the original highlight area. Following this, a logarithmic transformation was applied to adjust the contrast of the images, resulting in an elevation of grayscale values in previously dim areas, thereby yielding a set of High Dynamic Range (HDR) grating pattern images. Finally, the three-dimensional point cloud data of the measured object was obtained by performing phase calculations and parameter calibration. Results showed that this method can accurately calculate exposure time and frequency, and the fused raster images were clearer and more complete.For standard blocks, water pump pressure plate sleeves, and return disks with high-reflective surface, the point cloud reconstruction rates were 99.80%, 99.20%, and 99.40%, respectively, improving the measurement accuracy of the surface morphology of objects with high-reflective surface. In order to further verify the measurement accuracy of the proposed method, three standard blocks with different lengths, consistent width, and a height of 9 mm were stacked to form a standard step block. The height difference between two adjacent planes of the standard step block was measured using the proposed methodology, yielding a measured height difference of 9.099 7 mm. The relative error of the height difference was calculated to be +1.108%, indicating that the proposed methodology demonstrates a considerable level of measurement accuracy.