摘要
提出了一种有效的木材缺陷自动检测方法,试图赋予计算机从木材图像数据中自动辨别出缺陷的能力,主要分析了木材缺陷的形态、走向和分布规律.首先将要识别的木材图像变换到HSV色彩空间,分别对H,S和V层进行区域分割和Gabor小波变换,得到各个子图像块的局部区域的基于不同频率和方向的特征向量,用于描述高维的木材图像.接着将提取出的纹理特征归一化后送入SVM分类器,检测过程采用二次循环搜索方式,利用特征向量间的相似度进行缺陷的定位和识别.模拟实验结果表明,该方法可有效识别出缺陷区域,识别效果较好.
An effective method for automatic wood defects detection is proposed,and the computers are used to detect the defects automatically, which is mainly used for analyzing the distributions of defects. First, the original images are transformed to be detected into H, S and V color spaces,and cut them to many small blocks. And then Gabor wavelet is adopted to transform the sub-images into features on frequency and direction of local areas, which are used to describe the original high-dimensional wood images. Then the features are normalized and sent to the SVM classifier for classification. The two-layer search is adopted and the similarity between features is used to locate the defects. Many experiments show that the proposed method is effective and has a strong recognition ability to detect the wood defects automatically.
出处
《郑州大学学报(理学版)》
CAS
北大核心
2010年第1期93-97,共5页
Journal of Zhengzhou University:Natural Science Edition
基金
重庆邮电大学自然科学基金资助项目
编号A2009-35