田间原位光谱的鲜烟叶成熟度判别模型的研究

发布时间：2018-04-28 00:43

本文选题：可见光谱 + 光谱特征参数　；参考：《光谱学与光谱分析》2016年06期

【摘要】：在田间原位对烟叶成熟度进行判别,能够有效减少由于对成熟度判断错误而导致的烟叶损失率升高、质量下降的问题,而传统的人眼结合叶龄的田间成熟度判别方法缺少客观性,因此提出采用光谱特征参数结合支持向量机的方法对田间原位烟叶成熟度进行判别。以专家评定并在田间原位进行测量的五个成熟度等级共351个烟叶反射光谱作为试验样品,五个成熟度等级分别为M1,M2,M3,M4,M5。通过对反射光谱的分析发现,不同成熟度烟叶的光谱在可见光波段能够得到区分,而在近红外波段区分不明显,因此在可见光波段进行分析建模。分别采用可见光范围内的连续光谱(350~780nm)、特征波段(496~719nm)、光谱特征参数(绿峰幅值、绿峰位置、红边幅值、蓝边幅值、红边面积、蓝边面积、红边位置、蓝边位置)作为输入变量,采用支持向量机方法(supportvector machine,SVM)建立烟叶成熟度判别模型。结果表明,应用可见光光谱特征参数作为输入变量所建立的模型的正确识别率达到98.85%,而应用可见光连续谱、可见光特征波段作为输入变量的正确识别率分别为90.80%和93.10%。因此使用可见光光谱特征参数建立支持向量机的鲜烟叶成熟度判别模型对田间原位烟叶成熟度进行判别是可行的。
[Abstract]:In order to distinguish the maturity of tobacco leaves in the field, the loss rate of tobacco leaves increased and the quality of tobacco leaves decreased due to the wrong judgment of maturity. However, the traditional field maturity discrimination method based on human eye and leaf age is lack of objectivity, so the spectral characteristic parameters combined with support vector machine (SVM) is proposed to judge the maturity of field tobacco leaves in situ. A total of 351 tobacco leaf reflectance spectra of five maturity grades assessed by experts and measured in situ were used as experimental samples. The five maturity grades were M1OM2M3M3M4M5, respectively. Through the analysis of reflectance spectrum, it is found that the spectrum of tobacco leaves with different maturity can be distinguished in the visible light band, but not in the near infrared band, so the analysis and modeling are carried out in the visible light wave band. In this paper, the continuous spectral parameters (green peak amplitude, green peak position, red edge amplitude, blue edge amplitude, red edge area, blue edge area, red edge position, blue edge position) were used as input variables. Support vector machine (SVM) was used to establish the tobacco maturity discrimination model. The results show that the correct recognition rate of the model established by using the characteristic parameters of the visible light spectrum as input variables is 98.85, while the correct recognition rates of the visible light characteristic bands as input variables are 90.80% and 93.1010%, respectively. Therefore, it is feasible to use the characteristic parameters of visible light spectrum to establish the support vector machine (SVM) model to distinguish the maturity of fresh tobacco leaves in the field.
【作者单位】：中国农业大学现代精细农业系统集成教育部重点实验室;云南省烟草公司昆明市公司;云南省烟草农业科学研究院;
【基金】：国家自然科学基金项目(61144012) 中国烟草总公司云南省公司项目(2013YN17)资助
【分类号】：S572;O657.3

【相似文献】