基于高光谱成像技术快速无损测定花生中水分含量

Rapid and Non- destructive Determination of Moisture Content in Peanut Based on Hyperspectral Imaging Technology

花生中水分含量的高低直接影响花生及其制品的贮藏期,而现有的测定方法存在步骤多、时间长等问题。试验利用高光谱成像技术对花生中水分含量进行快速无损检测分析。通过采集120个花生样品的图像信息,从校正后的图像中提取花生目标区域的平均光谱作为花生光谱信息进行分析;同时,优选最佳的光谱预处理方法和建模方法建立花生中水分含量全波段模型,在此基础上利用回归系数法,确定重要波长并建立模型。结果表明,二阶导数(2nd-der)偏最小二乘法(PLS)全波段模型预测水分含量能力最佳,校正集和预测集的相关系数分别为0.91和0.84,标准偏差分别为0.28和0.38;回归系数法确定的14个波长所建简化模型的性能与全波段相当,校正集和预测集的相关系数分别为0.82和0.81,标准偏差分别为0.39和0.43。因此,高光谱成像技术可以快速无损测定花生中水分含量,其具有快速运算特点的重要波长模型可以更加方便地应用于花生加工产业中。

Moisture content directly affects the storage period of peanut and its products. The existing methods of determination of moisture content are more steps and time- consuming. In this paper, moisture content in peanut is rapidly and non- destructively detected by hyperspectral imaging technology（HSI）. Image information of 120 peanut samples is acquired,and average spectra of peanut of resign of interest are extracted from the corrected image as peanut spectral information to analyze. The whole wave- band model of moisture content in peanuts is established by selecting the best spectral pre- processing method and modeling method. Based on regression coefficient method,important wavelengths are identified and model is set up. The results show that 2^（nd）- der- PLS whole wave- band model had best ability to predict moisture content with Rc of 0.91 and SEC of 0.28,Rp of 0.84 and SEC of 0.38. The performance of the simplified model established by the regression coefficient is equivalent to that of the whole wave- band with Rc of 0.82 and SEC of 0.39,Rp of 0.81 and SEC of 0.43. The research show that HIS could quickly and non- invasively determine the moisture content in peanuts, and the important wavelength model with the characteristics of fast operation could be more convenient for the application in the peanut industry.