基于可见/近红外高光谱成像技术的鸡蛋新鲜度无损检测

以400~1 000nm高光谱系统获得鸡蛋样本的高光谱图像,利用蒙特卡洛法检测异常样本,采用不同预处理方法处理原始光谱;应用竞争性正自适应加权算法(Competitive Adaptive Reweighted Sampling,CARS)、遗传偏最小二乘法(Genetic Algorithms PLS,GAPLS)和间隔蛙跳法(Interval Random Frog,IRF)对预处理后光谱数据提取特征波长;分别建立基于全光谱和特征波长的偏最小二乘回归(Partial Least Squares Regression,PLSR)和最小二乘支持向量机(Least Squares Support Vector Machine,LS-SVM)鸡蛋新鲜度预测模型。结果表明:标准正态变量变换(Standardized Normal Variate,SNV)法为最优预处理方法;利用CARS、GAPLS和IRF分别选出8,35,74个特征波长;基于GAPLS提取的特征波长的LS-SVM模型最优,其校正相关系数(Rc)为0.899,预测相关系数(Rp)为0.832。表明基于高光谱成像技术的鸡蛋新鲜度无损检测是可行的。

Non-destructive determination of eggs freshness by Vis/NIR hyperspectral imaging technology

Hyperspectral images of egg samples were obtained using a hyperspectral system of 400~1 000 nm. The anomalous samples were detected by Monte Carlo method, and the original spectra were processed by three kinds of pretreatment methods. The characteristic wavelengths were extracted from the spectral data after pretreatment using competitive adaptive reweighed sampling (CARS), Genetic Algorithms PLS (GAPLS) and Interval Random Frog (IRF). The Partial Least Squares Regression (PLSR) and Least Squares Support Vector Machine (LS-SVM) eggs freshness prediction models based on full spectrum and characteristic wavelength were established respectively. The results showed that the standardized normal variate (SNV) method was the best pretreatment method. CARS, GAPLS and IRF were used to select 8, 35 and 74 characteristic wavelengths respectively. LS-SVM model based on characteristic wavelengths by GAPLS method was the best, and correlation coefficient of correlation (Rc) and prediction(Rp) of the model were 0.899 and 0.832 respectively. It was indicated that the non-destructive measurement for egg freshness based on hyperspectral imaging technology was feasible.