关于融合多针图确定物体三维表面绝对深度的研究  

ON FUSING MULTIPLE NEEDLE MAPS FOR RECOVERING 3-D ABSOLUTE DEPTH

作　　者：杨敬安[1] 

机构地区：[1]合肥工业大学人工智能所,合肥230009

出　　处：《自动化学报》1996年第1期58-67,共10页Acta Automatica Sinica

基　　金：国家自然科学基金资助高技术项目

摘　　要：利用ＰＳ（Ｐｈｏｔｏｍｅｔｒｉｃ Ｓｔｅｒｅｏ）系统很易确定物体表面方向及相对深度，但不能确定绝对深度．为确定绝对深度，本文提出的算法首先利用ＢＰＳ（Ｂｉｎｏｃｕｌａｒ Ｐｈｏｔｏｍｅｔｒｉｃ Ｓｔｅｒｅｏ）系统获得一对表面方向图．然后，基于ｇｅｏｄｅｓｉｃ ｄｏｍｅ分割这对方向图并计算图中对应区域间的视差．最后，通过施加多种约束，经适当融合及精确的视差匹配，确定景物物体３Ｄ表面绝对深度．这一方法对进一步研究怎样确定任意３Ｄ表面深度并复原景物结构有着十分重要的意义．The photometric stereo system can easily determine the surface orientation and relative depth of an object, but can not determine absolute depth. In order to determine absolute depth, this paper proposes an algorithm which obtains a pair of the needle maps between the corresponding areas in the two needle maps by binocular photometric stereo system, solves disparities between the corresponding areas in the two needle maps first and then fuses disparities correctly and match them accurately, finaly reconstructs 3 - D depth of the object surface. This method is very useful for as to conduct research on 3 - D surface depth reconstruction and to recover scene structure.

关 键 词：立体图象 多针图 SFS算法 智能机器人 

分 类 号：TP242.6[自动化与计算机技术—检测技术与自动化装置]