为解决结构光测量高动态范围表面物体时出现局部过度曝光或曝光不足的问题，提出一种改进的多曝光融合方法，利用自适应曝光代替手动曝光，并对图像融合过程进行优化。首先，将初始曝光时间下拍摄的图像利用直方图进行分析，将被测物体表面反射率不同的区域分为若干组，分别计算出每个组别的最佳曝光时间；在此基础上，拍摄不同组别对应最佳曝光时间下投射白光和条纹的图像，并去除图像中超过设定阈值的高灰度值区域，再将投射白光处理后的图像制作成掩模图，与相同曝光时间下投射条纹处理后的图像相乘，进而对多组相乘后的图像进行亮度压缩与融合；最后，通过CLAHE算法提高融合后所生成条纹图的对比度与清晰度，并对条纹解相后进行点云重建和尺寸测量。实验结果表明：文中方法中自适应曝光相较于手动曝光具有高效性和准确性，U型卡、连接块、圆盘三个高动态范围表面物体的点云重建率分别高达99.98%、99.74%、99.76%，测量出的标准块阶梯高度差绝对误差为0.062 mm，相对误差仅为0.69%，该方法有效解决了高动态范围表面物体测量时点云缺失的问题，提高了三维轮廓的测量精度。

For low dynamic range surface objects, a single exposure can provide sufficient exposure, but for high dynamic range surface objects, it is difficult to obtain high-quality fringe patterns with a single exposure, the multi-exposure fusion technology fuses the fringe pattern through multiple exposures, which can effectively improve the definition of the fringe pattern, thereby improving the accuracy of phase measurement. The traditional multi-exposure fusion technology needs to manually set the exposure time, which has problems such as low efficiency and poor exposure accuracy, in this paper, the method of adaptive exposure is used to obtain the exposure time, which avoids the disadvantages of manual exposure. Although the fringe image fused by traditional multi-exposure fusion technology has removed overexposure points, the overall quality of the fringe image is still not high, therefore, this paper improves the fusion process of multi-exposure images, a fringe map with better image quality is obtained. Firstly, the images taken under the initial exposure time are analyzed by histogram, the areas with different reflectance on the surface of the measured object are divided into several groups, the optimal exposure time of each group is calculated respectively; on this basis, the images of projected white light and projected stripes were taken under the optimal exposure time corresponding to different groups, after removing the high gray value area exceeding the set threshold in the image, then making the image collected when projecting white light into a mask image, and multiplied with the image acquired when the fringes were projected at the same exposure time, perform brightness compression and fusion on multi-group multiplied images; finally, improve the contrast and clarity of the fringe image generated after fusion through the CLAHE algorithm, thereby performing stripe unwrapping and point cloud reconstruction. The adaptive exposure used in this paper is more efficient and accurate compared to manual exposure (Fig.6), for the three high dynamic range surface objects of the U-card, the Connection Bock and the Disc, the fringe image fused by the method in this paper has no overly bright and overly dark areas, and the overall quality of the fringe pattern is good (Fig.7, Fig.8, Fig.9), there is no obvious loss in the point cloud\image after 3D reconstruction (Fig.10, Fig.11, Fig.12), the number of point clouds measured by the method in this paper is similar to the number of three-dimensional point clouds measured by the spray imaging agent method, the reconstruction rate has reached more than 99.5% (Fig.13), the measured absolute error and relative error of the standard block step height are only 0.062 mm and 0.69% (Tab.2). Aiming at the failure of 3D contour detection of high dynamic range surface objects, this paper proposes an improved multi-exposure fusion method, replacing manually setting exposure with adaptive exposure, at the same time, in the process of image fusion, the contrast and clarity of the fringe image are improved by setting the threshold, brightness compression, and using the CLAHE algorithm. Experimental results show that adaptive exposure is more efficient and accurate than manually setting exposure, the point cloud reconstruction rate of different high dynamic range surface objects is above 99.75%, the measured absolute error and relative error of the standard block step height are only 0.062 mm and 0.69%, effectively solved the problem of missing 3D point clouds for detecting high dynamic range surface objects, improved measurement efficiency and accuracy of 3D profile measurement.