支持向量机在烤烟烟叶自动分级中的应用

Application of Support Vector Machine in the Automatic Grading of Flue-cured Tobacco Leaf

烟叶自动分级是国内外烟草行业的重要研究课题之一。目前采用神经网络、模式识别等技术在对烟叶样本图像进行自动特征提取与分级时,分级的精度不很理想。在此,本文将支持向量机技术引入到烟叶自动分级中。实验表明,该技术可以为烟叶的自动模式识别提供稳定的参数值,与传统的神经网络方法相比,克服了固有的过学习和欠学习问题,并且对复杂模式的识别能力较强,已达到人类专家分级水平,为烟叶自动分级的研究开辟了新途径。

The Automatic Grading of Tobacco Leaf is an important research topic in the domestic and international tobacco industry. The current use of neural networks, pattern recognition and other technology in the image of tobacco samples for automatic feature extraction and grading, the grading accuracy is not very satisfactory. So Support Vector Machines is introduced in the Automatic Grading of Tobacco Leaf in the paper. It is concluded by experiments that this method can automatically provide a stable pattern recognition parameters, it overcomes the inherent defect of neural networks and achieves the standards of tobacco experts. This technique has become a new vehicle to the Automatic Grading of Tobacco Leaf.