利用局部二值模式和方向梯度直方图融合特征对木材缺陷的支持向量机学习分类
|
摘要
根据木材缺陷类型及视觉特点的不同,将木材缺陷分为6类,加上正常无缺陷木材共分7类;实验将图像样本转化为灰度图后生成680个训练样本数据集和94个测试样本数据集。在分析单一方向梯度直方图(HOG)特征及局部二值模式(LBP)采用不同核函数对木材缺陷分类性能的基础上,提出局部二值模式和方向梯度直方图融合特征对木材缺陷分类。融合特征经主成分分析并降维,利用支持向量机的4种不同核函数分别验证局部二值模式和方向梯度直方图融合特征对木材缺陷分类的性能。结果表明:利用局部二值模式和方向梯度直方图融合特征比单一缺陷特征具有更高效的分类性能;采用线性核函数及高斯核函数对局部二值模式和方向梯度直方图融合特征进行支持向量机分类,分类准确率分别可达98.9%和97.8%,木材缺陷可实现自动检测分类。
According to the different types of wood defects and visual characteristics, the defects were divided into seven types, together with normal non-defective wood. The experiment transforms image samples into gray images and generates 680 training sample data sets and 94 test sample data sets. With the analysis of the characteristics of single histogram of oriented gradient(HOG) and local binary pattern(LBP) using different kernels to classify wood defects, the fusion features of local binary pattern and histogram of oriented gradient were proposed to classify wood defects. By principal component analysis and dimensionality reduction, four different kernel functions of support vector machine were used to validate the performance of local binary pattern and histogram of oriented gradient fusion features for wood defect classification. The fusion feature has more efficient classification performance than single defect feature; the fusion feature of local binary pattern and directional gradient histogram is classified by support vector machine using linear kernel function and Gauss kernel function, and the classification accuracy can reach 98.9% and 97.8%, respectively; therefore, wood defect can be automatically detected and classified.
According to the different types of wood defects and visual characteristics, the defects were divided into seven types, together with normal non-defective wood. The experiment transforms image samples into gray images and generates 680 training sample data sets and 94 test sample data sets. With the analysis of the characteristics of single histogram of oriented gradient(HOG) and local binary pattern(LBP) using different kernels to classify wood defects, the fusion features of local binary pattern and histogram of oriented gradient were proposed to classify wood defects. By principal component analysis and dimensionality reduction, four different kernel functions of support vector machine were used to validate the performance of local binary pattern and histogram of oriented gradient fusion features for wood defect classification. The fusion feature has more efficient classification performance than single defect feature; the fusion feature of local binary pattern and directional gradient histogram is classified by support vector machine using linear kernel function and Gauss kernel function, and the classification accuracy can reach 98.9% and 97.8%, respectively; therefore, wood defect can be automatically detected and classified.
引文
[1] 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.
[2] 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.
[3] 张益翔.基于小波变换和局部二值模式的木材表面缺陷识别[D].南京:南京林业大学,2017.
[4] 傅红普,邹北骥.方向梯度直方图及其扩展[J].计算机工程,2013,39(5):212-217.
[5] 徐渊,许晓亮,李才年,等.结合SVM分类器与方向梯度直方图特征特征提取的行人检测[J].计算机工程,2016,42(1):56-60,65.
[6] 殷羽,郑宏,高婷婷,等.基于联合方向梯度直方图特征特征的车牌识别算法[J].计算机工程与设计,2015,36(2):476-481.
[7] 郭斯羽.面向检测的图像处理技术[M].长沙:湖南大学出版社,2015.
[8] 万源,李欢欢,吴克风,等.局部二值模式和方向梯度直方图特征的分层特征融合的人脸识别[J].计算机辅助设计与图形学学报,2015,27(4):640-650.
[9] 孙玉,刘贵全.基于方向梯度直方图特征与局部二值模式特征的人脸识别方法[J].计算机工程,2015,41(9):205-208,214.
[10] TAN X Y,TRIGGS B.Enhanced local texture feature sets for face recognition under difficult lighting conditions[J].IEEE Transactions on Image Processing,2010,19(6):1635-1650.
[11] MURALA S,WU Q M J.Local mesh patterns versus local binary patterns:biomedical image indexing and retrieval[J].IEEE Journal of Biomedical & Health Informatics,2014,18(3):929-938.
[2] 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.
[3] 张益翔.基于小波变换和局部二值模式的木材表面缺陷识别[D].南京:南京林业大学,2017.
[4] 傅红普,邹北骥.方向梯度直方图及其扩展[J].计算机工程,2013,39(5):212-217.
[5] 徐渊,许晓亮,李才年,等.结合SVM分类器与方向梯度直方图特征特征提取的行人检测[J].计算机工程,2016,42(1):56-60,65.
[6] 殷羽,郑宏,高婷婷,等.基于联合方向梯度直方图特征特征的车牌识别算法[J].计算机工程与设计,2015,36(2):476-481.
[7] 郭斯羽.面向检测的图像处理技术[M].长沙:湖南大学出版社,2015.
[8] 万源,李欢欢,吴克风,等.局部二值模式和方向梯度直方图特征的分层特征融合的人脸识别[J].计算机辅助设计与图形学学报,2015,27(4):640-650.
[9] 孙玉,刘贵全.基于方向梯度直方图特征与局部二值模式特征的人脸识别方法[J].计算机工程,2015,41(9):205-208,214.
[10] TAN X Y,TRIGGS B.Enhanced local texture feature sets for face recognition under difficult lighting conditions[J].IEEE Transactions on Image Processing,2010,19(6):1635-1650.
[11] MURALA S,WU Q M J.Local mesh patterns versus local binary patterns:biomedical image indexing and retrieval[J].IEEE Journal of Biomedical & Health Informatics,2014,18(3):929-938.