中文摘要：

基于立体视觉的物体深度图提取，在工业检测等许多方面均有较高的应用价值，但对于工程领域里的高光物体带来的误差，研究不多。为了较好地恢复具有高光反射的物体表面的三维深度信息，论文提出一种与立体视觉结合的多光源法去除高光的技术，使得后续的立体匹配更加可靠，深度计算更加准确。实验首先采用多光源法获取高光反射物体的多组图像，再经过图像预处理、摄像机标定、高光去除、立体匹配等步骤，结果证明这种方法有效地恢复了具有相对大面积的高光区域表面的物体的三维深度信息。

英文摘要：

3D depth retrieval from objects using computer stereo vision has a lot of applications in the industry.But few researches have been done respective to the objects with specular reflection.In this article,a new application of 3D depth retrieval of surfaces with specular reflection has been proposed.In this method,we have removed the highlight field of the metal surface using multi-lights illumination and correspondent algorithm,which makes the stereo matching more reliable and 3D depth calculation more accurate.To implement the method,we use one different light out of 4 each time to get the images.Then,after preprocessing,calibration,specular removal,stereo matching,the validity of the method is shown by the results.