基于高光谱漫透射成像可视化检测脐橙可溶性固形物

Visualized Detection of Soluble Solid Content Distribution of Navel Orange Based on Hyperspectral Diffuse Transmittance Imaging

厚皮类瓜果内部品质的无损检测是目前水果产业的检测技术瓶颈。本文采用高光谱漫透射技术对脐橙可溶性固形物(SSC)含量进行可视化分析研究。通过基线校正(Baseline)预处理结合连续投影算法(SPA)优选9个特征波长,建立SSC偏最小二乘回归(PLSR)模型,校正集相关系数r_(cal)为0.891,校正集均方根误差RSMEC为0.612°Brix,预测集相关系数r_(pre)为0.889,预测集均方根误差RMSEP为0.630°Brix。最后,计算各个像素点的SSC值结合图像处理技术得出SSC的可视化分布图,直观判断脐橙SSC含量高低。

Compared with the fruit with thin skin,it is more difficult to acquire the internal quality information of fruits with thick skin. In this study,the hyperspectral diffuse transmission technique was used to visually analyze the soluble solids content（ SSC） of navel orange. By comparison of the results,the model using the spectra pretreated by baseline correction as the input was the best one.Based on the baseline corrected spectra,successive projections algorithm（ SPA） was applied to select feature wavelengths and finally 9 bands were remained. The results of the partial least squares regression（ PLSR） model for SSC prediction indicate that the correlation coefficient of calibration（r（cal）） is 0. 891,the root mean square error of calibration（RSMEC） is 0. 612,the correlation coefficient of prediction（r（pre）） is 0. 889,and the root mean square error of prediction（RMSEP） is 0. 630,respectively. Using the spectra of feature wavelengths as the input,the multiple linear regression（ MLR） models for SSC prediction were calibrated. Based on the MLR model,each pixel value of the images was calculated. Combined with the image processing,the distribution maps of SSC in navel orange were drawn. So,the SSC of navel orange can be intuitive judged.