基于SVM的近红外黑木耳多糖含量分类

黑木耳作为一种胶质食用菌,富含多种营养物质,其中多糖是其中所占比重最大、含量最高的功能性成分之一。针对黑木耳多糖的分类,区别于传统的化学方法检测操作复杂、检测速度慢的问题,文中利用红外光谱技术,对东北黑木耳样本进行多糖成分进行无损检测,并利用支持向量机(SVM)的算法进行建模仿真。通过选定核函数的最优惩罚系数C=100、正则化系数V=10 ^-1,测得黑木耳多糖分类模型的识别结果的精确率为85.7%,召回率为87.3%,F1-score为0.864,达到了预期目标。试验证明,相比较于常规的黑木耳多糖检测方法来说,根据支持向量机算法建立的高精度黑木耳近红外光谱多糖分类模型是可行的。

Near-infrared Scanning Polysaccharide Content Classification of Auricularia Auricular Based on SVM

As a kind of colloidal edible fungus, auricularia auricular was rich in a variety of nutrients, among which polysaccharide was one of the most important functional ingredients. According to the classification of polysaccharides from Auricularia auricula, it was different from the traditional chemical detection method, and the detection speed was slow. In this paper, infrared spectrometry was used to carry out nondestructive detection of polysaccharides in northeast auricularia auricular samples, and the algorithm is modeled by SVM simulation. The accuracy of the recognition result of the auricularia auricular polysaccharide classification model was 85.7% and the recall rate was 87.3% , F1-score was 0.864, which achieved the expected goal. The experiment proved that compared with the conventional black fungus polysaccharide detection method, the high-precision auricularia auricular near-infrared spectrum polysaccharide classification model established by the support vector machine algorithm is feasible.