一种纹理与结构光叠加的立体视觉改进算法

为降低复杂纹理与结构光叠加对立体匹配算法视差结果的影响，提出一种针对纹理与空间编码结构光叠加情况下的立体视觉改进算法。通过4种空间编码结构光与三种立体匹配算法(BM、SGBM、NCC)对单一平面与电木压模工件进行视差重建，发现叠加会导致双目立体视觉无法计算出真实视差。本文通过对比引入结构光前后两次拍照视差，在纹理复杂区域内用加入结构光之前视差替换加入结构光后坏点视差，实现对纹理与结构光叠加处视差优选，从而提高视差重建完整性。实验对比发现，在四种不同性能结构光下改进前后的三维重建结果可以验证算法有效性，算法提升了平面区域深度检测精度的1%,优化了工件视差重建，达到了提高三维重建完整性的目的。

An improved stereo vision algorithm to solve the superposition of texture and structure light

Under the background of texture and spatial coded structure light superposition,this paper presents an improved stereo vision algorithm in order to reduce the influence of complex texture and structure light superposition on the parallax results of stereo matching algorithm.When a single plane and geometric work piece model images are reconstructed by four kinds of spatially coded structure lights and three stereo matching algorithms (BM,SGBM,NCC),we find that superposition phenomenon will lead to the deterioration of the real parallax calculation of binocular stereo vision.The parallax values of structure light image and unstructure light image are compared that the bad points with structure light are replaced by unstructure light image parallax in the complex texture area.It can optimize the parallax at the superposition of texture and structural light to improve the integrity of parallax reconstruction.A comparison of experiments show that,under four kinds of different coded structures,the effectiveness of the algorithm can be verified by the structure light image and unstructure light image of 3D reconstruction results.The accuracy of depth detection of plane region are improved by 1% by this algorithm,and the parallax reconstruction of work piece is optimized to improve the integrity of 3D reconstruction.