基于K均值算法的彩色编码条纹分色研究

Research on color distinction for color-coded stripe with K-means algorithm

在基于彩色编码结构光的三维重建中,由于受环境等的影响,拍到的条纹的颜色值已经不完全等于投射的颜色值,但每种颜色在RGB空间还是呈现一定的聚类性。基于这个特性,提出了基于K均值聚类的条纹分色方法。通过K均值聚类算法,对拍摄到的条纹图像进行颜色聚类,拟合出每一种编码颜色的一条直线作为模板,以此来实现了彩色条纹的分色。该方法鲁棒性较强,可以准确地将颜色区分开,为基于结构光的三维重建工作奠定了基础。

In 3D reconstruction based on color-coded structured light,because of the effect of environments,the colors of the captured stripes are not absolutely equal to the projected colors,but the captured colors still have the clustering characteristic.Based on this characteristic,a method of color-distinction with K-means algorithm was proposed.For every pixel＇s color of the captured stripe pattern,using the K-means clustering algorithm in RGB space,straight lines were fitted to form prototypes of these clusters,each line corresponding to one color,and then the straight lines were used to distinct the colored stripes.The experimental results show that the proposed method is robust,and it can correctly distinct the captured colors,laying the foundation for 3D reconstruction based on structured light.