基于高光谱成像的热加工过程鲍鱼含水量无损检测方法

Nondestructive Detecting Method for Moisture Content of Abalone during Thermal Processing Based on Hyperspectral Imaging

鲍鱼产品的含水量是其重要的质量参数,利用高光谱成像技术对不同热加工处理(水煮、微波和烤制)下鲍鱼含水量进行无损预测。采用卷积平滑(Savitzky-Golay smoothing, SG)和多元散射校正(Multiplicative Scatter Correction, MSC)两种方法分别对原始全波段光谱数据进行预处理,选取最优预处理方法;采用连续投影算法(Successive Projections Algorithm, SPA)和回归系数法(Regression Coefficient, RC)进行降维优选特征波长;建立基于全波段和特征波长的鲍鱼含水量偏最小二乘回归(Partial Least Squares Regression, PLSR)预测模型。结果表明,采用SG和SPA处理后的光谱数据建立的预测模型效果最优(RP2- = 0.9376, RMSEP = 4.63%, RPD = 3.85)。利用最优的预测模型对三种热加工下鲍鱼的含水量进行了可视化。研究表明高光谱成像技术可用于鲍鱼加工产品的无损质量监测。

The moisture content of abaloneproducts is an important quality parameter, and hyperspectral imaging technologywas used to make nondestructive prediction of abalone moisture content underdifferent thermal processing (boiling, microwaving, and baking). Savitzky-Golaysmoothing (SG) and Multiplicative Scatter Correction (MSC) were used topreprocess the raw full-band spectral data and select the optimal pretreatmentmethod, respectively;the Successive Projections Algorithm (SPA) and RegressionCoefficient (RC) were used to select characteristic wavelengths;the moistureprediction model of abalone was established using partial Least SquaresRegression (PLSR) based on the full wavelengths and characteristic wavelengths.The results showed that the model established by using spectral data processedby SG and SPA has the optimal performance (RP2- = 0.9376, RMSEP = 4.63%, RPD = 3.85). Theoptimal prediction model was used to visualize the moisture content of abaloneunder three kinds of heating treatments. The study showed that hyperspectralimaging techniques can be used for non-destructive quality monitoring ofabalone processed products.