基于四元数矩阵奇异值分解的木材缺陷检测分析

当前,木材彩色图像的缺陷检测主要是通过分离彩色空间的3个分量分别进行灰度处理,然后再合成为缺陷的图像.将基于RGB彩色空间的木材图像作为一个整体,提出四元数矩阵奇异值分解（QSVD）的木材缺陷检测.把RGB的彩色空间图像转换为四元数矩阵,利用四元数奇异值分解得到不同奇异值的特征图像,通过对特征图像的分析,得到不同的木材缺陷图像,并通过对奇异值特征图像的分析得到木材彩色图像的缺陷检测,并做分析.

Wood Defect Detection and Analysis Based on Quaternion Matrix Singular Value Decomposition

At present, the defect detection of wood color images which uses grayscale processing methods is mainly to separate the three components in the image color space respectively, and then synthesize the defect images. In this paper, the wood image based on RGB color space was treated as a whole, and a new wood defect detection algorithm based on quaternion matrix singular value decomposition （QSVD） was proposed. The RGB color space images were converted into the quaternion matrix space, and then the quaternion singular value decomposition was used to obtain the feature images with different singular values. With the analysis of the feature images, different wood defect images can be obtained. By linear weighting and non-linear weighting of singular value feature maps, the defect detection of wood color images were obtained and a brief analysis was made.