基于GLCM特征木材树种高光谱无损检测识别研究  

Research on hyperspectral nondestructive detection and recognition of wood species based on GLCM characteristics

作　　者：王辉[1] 李辉[1] 陈金阳 王敏 WANG Hui;LI Hui;CHEN Jin-yang;WANG Min(Panjin Vocational and Technology College, Panjin 124000, China)

机构地区：[1]盘锦职业技术学院

出　　处：《黑龙江科学》2019年第24期25-27,31,共4页Heilongjiang Science

基　　金：辽宁省百千万人才工程资助项目(2017014)

摘　　要：目前,木材树种无损检测识别技术已成为木材工业亟待解决的问题。提出了一种基于灰度共生矩阵(GLCM)纹理特征木材树种高光谱无损检测识别方法。在1000 nm^2500 nm的光谱范围内获取了样本高光谱数据,利用PCA对高光谱图像原始数据进行降维处理,得到主成分图像序列,并对主成分图像序列提取木材纹理特征参数,发现BP神经网络分类器的识别率为91.5%。实验表明:基于高光谱图像纹理特征能有效描述木材,用其实现对木材树种的识别是可行的,说明GLCM纹理特征参数对木材树种具有一定的描述能力,为木材树种的识别提供了新的思路。该方法能够较好地解决木材树种识别问题,对木材工业的现实生产、经营、管理等方面具有重要的现实意义。Wood is an indispensable raw material in people’s life.Non-destructive testing and identification technology of wood species has become an urgent problem to be solved in wood industry.In this paper,a non-destructive detection and recognition method for wood species based on gray level co-occurrence matrix texture feature is proposed.Sample hyperspectral data are obtained in the range of 1000-2500 nm.PCA is used to reduce the dimension of the original hyperspectral image data,and the principal component image sequence is obtained.The feature parameters of wood texture are extracted from the principal component image sequence,and the recognition rate of BP neural network classifier is reached to 91.5%.Experiments show that based on the texture feature of hyperspectral image,it can effectively describe wood,and it is feasible to recognize wood species with its realization.It shows that GLCM texture feature parameters have a certain description ability for wood species,which provides a new idea for wood species identification,and this method can better solve the problem of wood species identification,and it is of great significance to the actual production,management and management of wood industry.

关 键 词：GLCM 特征 木材树种 高光谱 无损检测识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]