Because of the same contour and luster for different types of machine-fried Longjing Tea, the intrinsic quality will be the pivotal factor in classification. Therefore, a combined technique with hyper spectral (HS) and support vector machine (SVM) was proposed in this paper to identify the class of intrinsic quality of machine-fried Longjing Tea. By using hyper spectral technology, the spectral feature parameters can be obtained, such as absorption depth, absorption area, red edge, red valley location and normalized difference vegetation index, etc.,within 350-2500 nm wavelength range. The correlation between these spectral feature parameters and tea classes was calculated. Based on the support vector classification theory with a penalty coefficient C, the key kernel functions and classification functions were identified by taking these spectral feature parameters as inputs. The identification model was constructed for classing Longjing Tea's classes. The model was also used in the classification experiment for different types of machine-fried Longjing Tea. The classification accuracy rate for machine-fried Longjing Tea's of this method is 98.3%, which proved that this method is feasible in machine-made Longjing Tea's classification.%随着机炒龙井茶外形和色泽趋向统一整齐,茶叶的内在品质相应成为评判等级的关键,由此,该文提出了一种高光谱与支持向量机分类相结合的技术,进行基于机炒龙井茶内在品质的等级识别方法研究.应用高光谱技术提取了龙井茶在350～2 500 nm波长范围内的吸收深度、吸收面积、红边位置、红谷位置、归一化植被指数等光谱特征参数,并对这些特征参数与茶叶等级的相关性进行了研究:然后利用带惩罚系数C的支持向量机分类理论,以光谱特征参数为输入量,分析确定了模型中关键的核函数和分类函数,构建了龙井茶等级识别模型,并进行了不同等级机炒龙井茶的分类识别验证.结果表明:采用所研究的方法和建立的模型对龙井茶进行等级分类准确率达到98.3％,证明应用该方法进行机炒龙井茶的等级分类识别是可行的.
展开▼
