基于图像融合的木板表面缺陷特征提取方法研究

Research on Feature Extraction of Defects on Wood Surfaces Based on Image Fusion

木材和实木家具表面在生产过程中有时会出现裂纹、凹点等缺陷,不同纹理背景和油漆反光会给缺陷识别带来很大困难。为了识别木板表面缺陷,通过光源对同一木板表面在4个不同角度照明并获取相应的4幅图像,组成图像序列,以获得更丰富的细节信息。提出一种基于主元分析法的图像序列融合方法,其融合了一组图像序列所包括的4幅图像的互补性信息,获取的融合结果可使缺陷特征更加明显。该方法引入了主元子空间之间的概念,可以在保留原有数据信息特征的基础上,提取主要信息。实验结果表明,基于主元分析方法的图像序列融合能更好地提取木板表面缺陷特征。所获得的特征图像可用于下一步对缺陷进行自动识别和分类。

There are sometimes defects such as cracks,Bump,etc on the wood surfaces in the production process of wood and furniture.The different texture of the wood surfaces and the reflex light of the varnished surfaces enhance the difficult of defects detection on Wood Surfaces.In order to inspect the defects,we illuminated the surface of wood from four different angels and gained the respective four images for more detail information.A fusion method for the image series based on principal component analysis（PCA） was presented in this paper,which fuses the complementary information of the image series including the four images and gets a result with more distinct defects.This method introduces a principal component subspace and reserves original information when extracting mainly information.The emulated results show that more distinct features can be extracted from the four images of a same surface by fusing the image series with PCA.The extracted features can be used to automatically detect and classify the defects on the wood surfaces in the future task.