Generally,even the subtle differences of wood textures among different species can be clearly identified with their scanning electron microscope( SEM) micrographs. Therefore,the features extracted from the SEM micrographs can provide a reliable basis for the classification of wood micrographs. In this paper,a novel method of wood SEM micrographs was proposed for feature extraction based on Graph Cuts. Firstly,the wood SEM micrographs would be enhanced before partitioning the images into multiple segments by Graph Cuts,then extracting the contours of wood tissues and smoothing the contours of objects by mathematical morphology. Finally,nine kinds of geometric and statistical features and their computing methods were given. The experimental results on 12 wood species showed that the features we proposed had a great difference among the given wood species( also in the earlywood and latewood of each given species) ,which provided an important basis for the wood assortment recognition.%提出一种基于 Graph Cuts图像分割的木材 SEM 图像特征提取方法。首先对木材 SEM 图像增强，应用Graph Cuts分割图像；提取木材组织轮廓后，使用数学形态学平滑对象轮廓；最后提出木材细胞的9种几何和统计特征及其计算方法。针对12个木材树种的试验结果表明：本方法给出的特征在给定树种及各树种的早、晚材间均具有较大的差异，为木材材种识别提供了重要依据。
展开▼
