基于多目标遗传算法的管孔组合特征识别

提出一种管孔组合方式的自动识别方法.提取木材显微横切面的边缘后,基于类圆区域面积直方图,通过最大类间方差将封闭区域分为2类,将导管与其他形状相似、大小不同的组织分开；同时,引入封闭区域平均面积作为另一目标函数.最后采用管孔集邻接度来判断管孔的组合方式.试验表明该方法能够对各种导管的分布和组合获得满意的分割效果,并给出管孔的组合方式.

Multi-Objective Genetic Based Pore Combination Recognition

This paper proposes an automatic method of pore combination recognition,which is an important feature to hardwood recognition.After extracting edge from wood microscopic cross-section,based on area histogram of the similar circle regions,the method classifies all regions into two classes with maximum between-class variance,so as to distinguish the pore from other textures,which are similar in shapes but different in sizes.Meanwhile,second objective function about average area of closed regions is used to improve the pore segmentation performance.At last,the method uses adjacency degree of pore set to judge pore combination.The experiments demonstrate that the task of pore segmentation can be completed successfully for all kinds of pore distribution and combination,and also the correct combinations of pores are given.