面向交叉点网络的符号解码方法的研究

计算机视觉是近年来研究的热点，结构光视觉是双目计算机视觉的一个分支，该技术采用一台投射器代替双目立体视觉的一台摄像机，由于其投射的结构光具有编码唯一性的特点，从而解决了双目视觉中像素匹配的难题。结构光技术中解码成为了一项重要的工作，根据符号的特点提出一种基于交叉点网络的符号解码方法。该方法有效地实现了交叉点的提取，并形成全局的交叉点网络。针对细化的十字交叉变形，研究了交叉点合并方法，提出了基于识圈法的内部交叉点定位方法，以及基于向量角度的符号分类方法，有效地实现了符号码字的还原。最终实现物体的三维重建目标。

Study on symbol decoding method based on cross points net

Computer vision has been the hot spot for several decades.Structured light technique is a branch of binocular computer vision by utilizing a projector instead of one camera in binocular stereovision.Because the projector structured light pattern owns the characteristic of＇ code uniqueness, structured light technique is able to solve the pixel matching problem in binocular vision.The symbol decoding becomes an important job in structured light technique,and according to the symbols＇ property,a decoding method based on cross points net is proposed.This method extracts the cross points effectively and creates the global cross points net.Aiming at the phenomena of cross chiasm distortion,a cross point combination approach is studied, and an inner cross points location method based on circular recognition algorithm and symbols classification approach based on vector angle algorithm are proposed.The proposed approaches have effective performance on symbol codeword retrieving.Finally,the target of object＇s 3D reconstructed is realized.