中文摘要：

在随机光场的辅助照射下,综合利用图像信息和几何约束信息,对匹配施加更多的约束,在匹配中即可以得到三维重建结果,从而使立体图像匹配和三维点云生成过程融为一体;提出了一种带权值的匹配窗口,在一定程度上改善了匹配效果;基于连续性约束,提出了用生长法给出初始匹配,可以极大地减少算法迭代次数,提高匹配速度。为了能够进一步提高算法的实用性,还讨论了灰度矫正处理方式对算法的影响。在实验中,完成了实际物体的三维表面点云重建,并与典型商用系统进行了重建结果的对比分析,验证了本文算法在精度和三维重建效果方面的优越性。

英文摘要：

With the projection of random illumination, a least square method is used on the knowledge of image grey intensity information as well as binocular camera parameters. The binding of camera parameter information gives more constraints on the matching and makes the simultaneousness of matching and reconstruction possible. The employment of the weighted matching window makes the matching results more accurate. A region growing scheme from the continuity constraint is exploited to give good initial matching parameters to decrease the iteration time distinctively and make the algorithm converge rapidly. Different radiometric correction methods are compared and discussed. In the reconstruction experiments, the real object is virtually reconstructed, and the 3-D cloud is compared with the typically commercial measurement system. The result demonstrates the validation and the effectiveness of the algorithm.