摘要
实现木材机械加工表面质量自动评价对保证木制品质量和提升其价值有重要意义。根据ASTM_D_1666-87R04标准中规定的测试方法,对樟子松、白桦、水曲柳板材进行压刨和砂光加工试验,得到有Raised grain、Fuzzy grain、Chip marks缺陷和无缺陷的木材试样。扫描木材试样以获取其表面图像,计算图像的直方图纹理特征和小波纹理特征。对每种特征分别建立基于BP神经网络的分类器并进行训练和测试。测试结果表明,小波纹理特征分类器的正确率达到91.3%,其评价效果优于直方图纹理特征分类器。
Automated evaluation of machining wood surface quality plays an important role in the overall quality and value of wood products. According to the test methods of ASTM_ D_ 1666- 87R04,this paper practiced planning and sanding tests on Mongolian Scotch Pine,Asian White Birch and Manchurian Ash lumbers. Raised grain,Fuzzy grain,Chip marks and defect free samples were obtained. The surface images of the sample were captured by scanning. Histogram texture features and wavelet texture features were retrieved from those surface images. Two BP ANNs were developed and tested according to these two kinds of features separately. The test results showed that the wavelet texture feature classifier achieved an accuracy of 91. 3% which is much better than the performance of histogram features classifier.
出处
《森林工程》
2015年第1期59-62,共4页
Forest Engineering
基金
黑龙江省博士后基金资助项目(LBH-Z09283)
关键词
木材加工
表面质量
自动评价
小波纹理分析
wood machining
surface quality
automated evaluation
wavelet texture analysis
