基于量化构造特征参数的树种计算机识别算法

Study on the Mold of Wood Computer Identification System Based on Quantified Anatomy Properties

简述了以往木材树种识别专家系统的数学算法,探索提出了一种新的基于由图像处理技术提取的木材构造解剖特征量化参数的计算机木材识别的算法———最大相似法 该方法可采用最小差值判别、树种综合特征阈值及综合加权相似3种方式来达到木材树种识别的目的,同时讨论了这3种方式的计算方法 这三者的主要区别在于:最小差值判别法是一种绝对的比较方法,它不考虑提取量化特征过程中所存在的误差与树种自身存在的变异,主要依靠多参数来修正误差;树种综合特征阈值法是一种相对的比较方法,它承认了客观误差的存在,利用适当的阈值来修正测量误差与树种品质的变异;综合加权相似法是在主成分综合分析的基础上,将每个特征量对树种的影响程度大小(即贡献率的大小)作为系数进行相似系数的计算,而前两者则将所有特征量视为平等 通过以上3种方式基于量化参数实现的木材树种识别方式具有比传统的专家系统检索方法更为客观。

The older mathematic arithmetic of wood identification system is simply described in this paper.A new identification arithmetic,that is the most similar method,is described.The new arithmetic of wood identification system is based on the wood quantified anatomic characters. The quantified characters are gotton by the digital image process. The identification arithmetic can be realized by the discriminance of minimal difference of parameters, the discriminance of limen of tree comprehensive character and similarity method of comprehensive. At the same time, the calculation process of the three methods is discussed. The main difference of the three methods is: the first is comparing absolutely, and it′s error is by corrected more parameters. The second is comparing relatively, and it′s correcting error by suitable limen. The third is based on the principal component analysis, and computer the similarity coefficient according to component score coefficient matrix. But the parameters is considered as same by the discriminance of minimal difference of parameters and the discriminance of limen of tree comprehensive character. The new mold is more objective and affected less by error than the older wood identification system.