论文部分内容阅读
颜色和部位是烟叶分级的重要组成部分,是影响烟叶质量的关键因素。以红塔集团提供的6类工业分级烟叶标样作为试验样品,其中包括上(B)、中(C)、下(X)三个部位,每个部位包含桔黄(O)、柠檬黄(L)两个色组。采用基于主成分及Fisher准则(PPF)的方法和支持向量机(SVM)方法分析烟叶可见-近红外高光谱的颜色和部位特征,结果表明,采用PPF投影模型法分别对烟叶颜色、部位以及颜色和部位分组进行投影和相似性分析,两种颜色能完全区分,其相似度值为-1.00 08;上部烟叶和下部烟叶能完全区分,与中部烟叶有部分交集,其中上部烟叶和下部烟叶的相似性值为-0.405 3;6类分组烟叶样品能完全区分,且投影位置关系符合实际的烟叶外观特点。采用SVM方法分别对烟叶颜色、部位以及颜色和部位分组进行判别分析,烟叶颜色的平均识别正确率达到98%,部位的平均识别正确率为96%,颜色和部位分组的平均识别正确率为94%,判别效果良好。因此,应用可见-近红外高光谱分析烟叶的颜色和部位特征具有可行性,为烟叶质量评价、计算机辅助分级以及烟叶智能收购等方面提供参考,同时也为其他农产品外观特性的分析提供了一种新方法。
The color and location are important components of tobacco leaf classification and are the key factors affecting the quality of tobacco leaf. Six types of industrial grade tobacco leaves samples provided by Hongta Group were used as test samples, including three parts of upper (B), middle (C) and lower (X), each containing orange (O), lemon yellow Two color groups. The color and site features of visible-near-infrared hyperspectral spectra of tobacco were analyzed by principal components and Fisher criterion (PPF) and support vector machine (SVM). The results showed that the color, location and color of tobacco And the location of subgroups for projection and similarity analysis, the two colors can be completely distinguished, the similarity value of -1.00 08; upper leaves and lower leaves can be completely distinguished, and the central part of the intersection of tobacco, of which the upper and lower leaves of tobacco similar The value of -0.405 3; 6 groups of tobacco samples can be completely distinguished, and the projection of the location of the tobacco leaf in line with the actual appearance characteristics. Using SVM method to discriminate the color, position, color and location of tobacco leaves respectively, the average recognition accuracy of tobacco leaf color was 98%, the average recognition accuracy of part was 96%, and the average recognition accuracy of color and part group was 94 %, Discriminant effect is good. Therefore, the application of visible-near-infrared hyperspectral analysis of the color and location characteristics of tobacco leaves is feasible and provides a reference for quality evaluation of tobacco leaves, computer-assisted grading and smart acquisition of tobacco leaf etc. It also provides a reference for the analysis of appearance characteristics of other agricultural products new method.