现有的应力波木材检测仪只能测定木材内部是否存在缺陷,无法对木材缺陷类型进行分类。笔者提出了一种结合应力波无损检测技术和支持向量机(SVM)的木材缺陷识别分类方法,该方法首先测量木材内部的应力波传播速度,以此作为分类特征,利用支持向量机对木材的内部缺陷进行分类。为了验证该方法的有效性,选取健康的以及含有不同缺陷的山核桃木试样31件、松木试样28件,采集山核桃木试样应力波传播速度数据117组、松木试样应力波传播速度数据80组,以应力波传播速度为分类特征,利用支持向量机对木材的缺陷类型进行分类。结果表明:山核桃木试样缺陷分类准确率达到93.75%,松木试样缺陷分类准确率达到95%。该方法不仅能识别木材内部是否存在缺陷,还能对木材的空洞、裂缝、腐朽等缺陷进行准确分类。
Abstract
The existing stress wave testing can only determine the existence of defects in the wood, but can not classify the type of wood defect. This paper presents a method which combines stress wave nondestructive testing technology and support vector machine(SVM)to identify and classify wood defects. This method measures stress wave velocity in the wood firstly, and then classifies the internal conditions of wood using SVM with the stress wave velocity as the classification feature. In order to demonstrate the effectiveness of the proposed method, 31 pecan wood samples and 28 pine wood samples with different conditions were selected as experimental samples. The Arbotom detector from Rinntech Company in German was used to collect 117 groups of data of stress wave velocity from pecan wood and 80 groups of data of stress wave velocity from pine wood. The classification accuracy of pecan wood and pine wood are 93.75% and 95% respectively. This detection method can not only recognize wood defect but also can accurately distinguish the defect type including voids, cracks, and decay.
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 冯海林,李光辉,方益明,等.应力波传播模型及其在木材检测中的应用[J].系统仿真学报,2010,22(6):1490-1493.Feng H L, Li G H, Fang Y M, et al. Stress wave propagation modeling and application in wood testing[J].Journal of System Simulation, 2010,22(6):1490-1493.
[2] 杨学春,王立海.木材应力波无损检测研究[M].北京:科学出版社,2011.
[3] 王欣,申世杰.木材无损检测研究概况与发展趋势[J].北京林业大学学报,2009,31(1):202-205.Wang X, Shen S J. Advances in non-destructive testing for lumber[J].Journal of Beijing Forestry University, 2009,31(1):202-205.
[4] 业宁,王厚立,徐兆军,等.基于支持向量机的木材缺陷识别[J].计算机应用与软件,2006, 23(4):3-5.Ye N, Wang H L, Xu Z J, et al. Recognition of wood defection based on support vector machine[J].Computer Applications and Software,2006, 23(4):3-5.
[5] 范宇,张冬妍,孙丽萍,等.基于SVM的木材干燥过程含水率软测量研究[J].森林工程,2008,24(4):27-29.Fan Y, Zhang D Y, Sun L P, et al. Soft sensor for wood moisture content based on support vector machine[J].Forest Engineering, 2008,24(4):27-29.
[6] 宋世全.基于支持向量机的原木树种及原木内部孔洞缺陷辨识[D].哈尔滨:东北林业大学,2012.Song S Q, The identification of log species and hole defect in log based on the support vector machine[D].Harbin:Northeast Forestry University,2012.
[7] Pham D T, Muhamad Z, Mahmuddin M, et al. Using the bees algorithm to optimise a support vector machine for wood defect classification[C]//Memorias del. Innovative Production Machines and Systems Virtual Conference. Cardiff:IPROMS, 2007.
[2010-07-06]. http://repo.uum.edu.my/154/.
[8] Gu I Y H, Andersson H, Vicen R. Wood defect classification based on image analysis and support vector machines[J]. Wood Science and Technology, 2010, 44(4): 693-704.
[9] Li G, Wang X, Feng H, et al. Analysis of wave velocity patterns in black cherry trees and its effect on internal decay detection[J]. Computers and Electronics in Agriculture, 2014, 104: 32-39.
[10] Fan R E. Working set selection using the second order information for training SVM[J]. Journal of Machine Learning Research, 2005(6): 1889-1918.
[11] 张召,业宁,业巧林.基于纹理提取和SVM技术的自动木材缺陷识别[J].计算机工程与应用,2009,45(23):219-223.Zhang Z, Ye N, Ye Q L. Automatic wood defects recognition based on texture extraction and support vector machine technology[J].Computer Engineering and Applications,2009,45(23):219-223.
[12] Collobert R, Bengio S. SVM Torch: a support vector machine for large-scale regression and classification problems[J]. Journal of Machine Learning Research, 2001(1):143-160.
[13] 刘波,郝志峰,肖燕珊,等.交互迭代一对一分类算法[J].模式识别与人工智能,2008, 21(4): 425-431. Liu B, Hao Z F, Xiao Y S, et al. Alternating iterative one-against-one algorithm[J]. Pattern Recognition and Artificial Intelligence, 2008, 21(4): 425-431.
[14] 张学工.关于统计学习理论与支持向量机[J].自动化学报,2000, 26(1): 32-42. Zhang X G. Introduction to statistical learning theory and support vector machines[J]. Acta Automatica Sinica, 2000, 26(1):32-42.
[15] 史峰,王小川,郁磊,等.MATLAB神经网络30个案例分析[M].北京:北京航空航天大学出版社,2010.
基金
收稿日期:2014-08-28 修回日期:2014-12-22
基金项目:国家自然科学基金项目(61272313, 61302185, 61472368); 浙江省科技厅项目(2012C21015,2013C31018,2013C24026,2014C31044); 浙江省自然科学基金项目(LQ13F020013); 浙江省大学生创新创业孵化项目(2013R412055); 浙江省林业智能监测与信息技术研究重点实验室资助项目(100151402)
第一作者:王再超,硕士生。*通信作者:李光辉,教授,博士。E-mail: lgh@zafu.edu.cn。
引文格式:王再超,李光辉,冯海林,等. 基于应力波和支持向量机的木材缺陷识别分类方法[J]. 南京林业大学学报:自然科学版,2015,39(3):130-136.
PDF(2094840 KB)
