基于高斯-马尔可夫随机场的木材表面缺陷类型识别  

Identification of Wood Surface Defects Types Based on Gause-Markov Random Field

作　　者：白雪冰[1] 祝贺[1] 张庭亮[1] 王再尚 张娜[1] 

机构地区：[1]东北林业大学机电工程学院,哈尔滨150040 [2]哈尔滨电工仪表研究所,哈尔滨150040

出　　处：《森林工程》2013年第6期56-58,共3页Forest Engineering

基　　金：黑龙江省自然科学基金项目(C201208);黑龙江省博士后基金项目(LBH-Q10160)

摘　　要：为了识别死节、活节和虫眼三种木材表面缺陷类型,本文采用高斯-马尔可夫随机场模型提取木材表面缺陷图像的纹理参数,结合缺陷区域的矩形度和伸长度两个几何特征,形成14维特征向量。设计三层BP神经网络来识别缺陷的类型。试验表明,三种缺陷的整体识别正确率达到96.67%,验证了该方法的有效性。To identify the type of three wood surface defects： wormhole, dead knot, and live knot, Gause-Markov Random Field model is used to extract texture characteristics parameters of the defect area. In combination with the shape characteristics parameters of the defects area, the input feature vector of mode classifier is established. BP-ANN is designed to identify the type of defects. The experiment results show that the identification accuracy rate of wood surface defects could reach 96. 67%. The method can identify the type of wood surface defects successfully.

关 键 词：木材表面缺陷 高斯-马尔可夫随机场 BP神经网络 

分 类 号：S781.1[农业科学—木材科学与技术]