【作者】 刁宏志；

【导师】 曹军；

【作者基本信息】 东北林业大学 ， 机械设计及理论， 2008， 博士

【摘要】 随着我国经济规模和科技水平的飞速发展,木材需求量和应用范围不断扩大,木材的研究从宏观到微观不断深入,对木材细胞的数字化描述也已经成为木材研究的热点和重要方向之一。木材是典型的各向异性、天然孔型材料,它是由无数不同形状、不同大小、不同排列方式的细胞组成的。由于树木生长不均匀,使各种树木的木材构造具有多样性,不同细胞体、细胞密度和细胞形状导致了木材材性和物理力学性能存在较大差异,木材细胞的变化影响着木材性质的变化,木材的性质又决定木材经济价值,并直接影响木材的加工和利用,由此对木材细胞结构分析研究就变得尤为重要。本文在参阅国内外有关木材细胞研究的基础上,通过收集连续的木材细胞切片,得到了从髓芯到边材完整的年轮细胞图像;分析了木材细胞微观结构,通过试验确定了细胞面积和细胞胞壁厚度直接或间接影响木材的使用价值和加工利用性能。在对比几种传统的边缘算子对木材细胞图像分割效果缺陷的基础上,采用了水平集方法(Level SetMethod)完成对木材细胞图像进行分割处理,并采用了一种快速算法——窄带算法(Narrow Band),解决了水平集方法图像处理时间过长的问题,采取了改进的基于Mumford-Shah模型的水平集方法对木材细胞图像进行分割,避免了分割存在盲目性,此方法极大改善了分割的效果。本文提出了基于活动轮廓模型(Snake),对细胞的轮廓形态进行了识别处理。利用改进的Snake两步膨胀模型对细胞轮廓进行提取,同时应用贪婪优化算法,减少程序运行时间,在完成对木材细胞内腔和边缘分割的基础上,完成对木材细胞壁的识别。在准确分割的基础上,逐个得到了每个年轮的数据参数,在MATLAB环境下,基于最小二乘原理对针叶材细胞胞壁厚度和细胞面积进行了曲线拟合,建立了针叶材细胞年轮方向上细胞和胞壁厚度的分布规律模型,并与三次样条函数曲线拟合做了对比。由于六角形木材细胞显微图像,可以代表木材细胞的大部分的形状特征,本文完成了对规则正六角形、规则长六角形、按早晚材变化的六角形木材细胞显微图像的仿真。分析利用图像数据库的停用和关键技术。本研究将最小二乘原理、数字图像分割、数字图像处理、数字图像识别应用于木材科学研究,分析了从年轮中心到木材边缘的年轮的木材细胞图像,应用改进的水平集方法准确分割细胞图像,应用活动轮廓模型完成了对木材细胞的识别,并完成了木材细胞六角形仿真,对木材材性的深入研究等领域中有重要的潜在应用价值,对提高木材对微观性能认识,扩大木材的适用范围,具有重要意义。更多还原

【Abstract】 With rapid development of our national economic scale and scientific and technological level,the demand of wood and its application range is enlarged continuously,and the research on wood becomes more and more thorough and deepens from macroscopic to microscopic.At the same time,the digitized description for wood cell has become focus and an important direction of the research on wood.Wood is the typical anisotropic material with natural hole, which is composed of the cells with different shape.Different cells,cell density and cell shape lead to comparatively great difference in wood properties and physical mechanical properties, change of wood cells affects change of wood properties including wood quality,wood property and all other aspects,and wood properties decide economic value and directly affect manufacturing and utilization of wood,therefore,the research and analysis on wood cell structure become especially important.B ased on concerned domestic and foreign research on wood cells,through making sections of wood cells,intact cell images of annual ring was obtained;The microstructure of wood cells was analyzed by experiments,and the experiment results make certain that cell area and cell wall thickness affect on use value as well as processing and utilization performance directly or indirectly.Direct towards the defect of image segmentation effect that several traditional edge operators have on wood cells,level set method was employed to do the segmentation treatment for wood cell image in this paper and fast algorithm of level set method, narrow band algorithm,was adopted,which solved the problem of too ling image treatment time,and modified level set method based on Mumford-Shah model was employed to perform segmentation for wood cell image,which avoided blindness existing in segmentation and improved the effect of segmentation greatly.This paper brought forward that identifying processing of cell morphology should be performed based on active contour model(snake). The cell outline was collected with two-step expansion model of modified Snake,at the same time;greedy algorithm was applied to reduce the operative time of procedure.Upon the basis that the segmentation of wood cell intracavity and edge was finished,wood cell wall was identified.Based on precise segmentation,the digital of parameters of annual ring was obtained gradually,under MATLAB environment and based on least square principle,the curve fitting was carried out for cell wall thickness and cell area of coniferous wood,the distribution law model of cells and cell wall thickness on annual ring direction was established and compared with cubic spline functions fitting.Because sexangle microscopic image of wood cells can represent most shape features of wood cells,this paper had accomplished the simulation of microscopic image for wood cells of regular normal hexagon,regular long hexagon changeable hexagon as per Early and Late.The expression and storage of image database for wood cell images was finished by analyzing and using the discontinuation and key technology of image database.Principle of least squares,digital image segmentation,digital image treatment and digital image identifying were applied into the research of wood science in this research,the wood cell image of annual ring from annual center to wood edge was analyzed,the modified level set method was applied to segmentate cell image precisely for the first time,the identifying of wood cells and hexagon simulation for wood cells were accomplished with active contour model,which had important potential application value in deepening the research of wood properties to a further step,moreover,had an important significance for improving comprehend on wood micro-properties and enlarging the applicable range of wood.更多还原