基于灰度共生矩阵和模糊BP神经网络的木材缺陷识别
作者：牟洪波 王世伟 戚大伟 倪海明
来源：《森林工程》2017年第04期
        摘 要：
        针对当前木材资源紧缺的严重形势，提高木材缺陷检测的准确率显得尤为重要。利用X射线无损检测技术获取木材缺陷的图像，并且通过灰度共生矩阵的方法能够有效地提取图像的主要特征值即特征向量，同时将模糊数学与BP神经网络相结合设计出模糊BP神经网络（FBP），并采用最大隶属度方法对特征向量进行模式识别，从而实现木材缺陷的自动识别和分类。经多次学习训练，结果表明FBP网络的平均识别成功率在90%以上。因此，FBP神经网络对木材缺陷有较高的识别准确率，可以为缺陷识别提供重要的理论依据。
        关键词：
        木材缺陷；灰度共生矩阵；特征提取；模糊BP神经网络
        中图分类号：S 781.5 文献标识码：A 文章编号：1001-005X（2017）04-0040-04
        Wood Defects Recognition Based on Gray-level Co-occurrence Matrix and Fuzzy BP Neural Network
        Mu Hongbo，Wang Shiwei，Qi Dawei*，Ni Haiming
        （College of Science，Northeast Forestry University，Harbin 150040）
        Abstract：
        It is important to enhance the accuracy in wood defects detection against the serious shortage of wood resources situation.Wood defects images were acquired by X-ray nondestructive testing technology.Feature vector which was the major characteristics of images could be effectively extracted by gray level co-occurrence matrix.At the same time，the fuzzy BP neural network（FBP）was designed by the combination of fuzzy mathematics and BP neural network.The maximum membership degree principle was used to do the pattern recognition of feature vectors，and then the automatic recognition and classification of wood defects could be realized.After a lot of training，results showe
d that the average recognition rate of FBP is above 90%.Therefore，FBP has a high recognition accuracy for wood defects，which can provide an important theoretical basis for defects identification.

更多推荐

缺陷,木材,方法,检测,识别,灰度