In this study,the hyperspectral (900～1700 nm)imaging technique combined with chemical statistical algorithms was used to study the non-destructive testing about egg's haugh unit.The original spectral information of the samples was collected and the PLSR model was established after S-G,MSC,SNV pretreatment to choose the best preprocessing method.The PLSR and PCR models were established,they based on the characteristic wavelengths screened by using the weighted β coefficients of PLSR and the genetic algorithm GA.To compare the models' effect and select the best extraction method of characteristic wavelength and modeling method.The results showed that the modeling effect were not improved after preprocessing.In other words,using the original spectrum as the modeling object,the PLSR model was optimized by using genetic algorithm to extract the characteristic wavelength,the RC was 0.8118,RMSEC was 2.9677,RP was 0.8203,RMSEP was 3.2762.It can be concluded that the use of hyperspectral techniques for nondestructive testing of eggs in haugh unit was feasible.At the same time,it provided a theoretical basis for the rapid identification and classification of fresh eggs.%本实验利用高光谱(900～ 1700 nm)成像技术结合化学统计学算法,对鸡蛋哈氏单位的无损检测进行研究.采集样品原始光谱信息,经S-G卷积平滑、MSC、SNV预处理后建立PLSR模型,优选最佳预处理方法;利用PLSR的加权β系数、遗传算法GA提取特征波长,分别建立PLSR、PCR模型,比较建模效果,优选最佳特征波长提取方法及建模方法.结果表明,预处理后建模效果并未提高,即以原始光谱作为建模对象,采用遗传算法提取特征波长后建立的PLSR模型效果最好,其RC为0.8118,校正集均方根误差是2.9677,RP为0.8203,预测集均方根误差是3.2762.因此,利用高光谱技术无损检测鸡蛋哈氏单位是可行的,同时为鸡蛋的新鲜度快速判别、分级分选提供理论依据.
展开▼
