基于漫反射高光谱成像技术的哈密瓜糖度无损检测研究

利用高光谱成像系统获得网纹类哈密瓜糖度漫反射光谱信息，选择有效波段500～820 nm进行哈密瓜糖度检测建模回归分析。对比了多元散射信号修正和标准正则变换校正方法，原始光谱、一阶微分、二阶微分光谱预处理方法对建模精度的影响;采用偏最小二乘法、逐步多元线性回归和主成分回归方法对比分析了带皮哈密瓜和去皮哈密瓜糖度检测模型效果。结果表明，对原始光谱经过MSC和一阶微分光谱处理后，采用PLS和SMLR方法均可取得很好的建模效果，应用PLS法检测带皮哈密瓜糖度是可行的，其校正集相关系数(R_c)为0.861，RMSEC为0.627，预测集相关系数(R_p)为0.706，RMSEP为0.873;应用SMLR法检测去皮哈密瓜糖度效果最佳，校正集相关系数(R_c)为0.928，RMSEC为0.458，预测集相关系数(R_p)为0.818，RMSEP为0.727。研究表明，应用高光谱成像技术检测哈密瓜糖度具有可行性。

Nondestructive Measurement of Sugar Content of Hami Melon Based on Diffuse Reflectance Hyperspectral Imaging Technique

The research on nondestructive test for detecting the sugar content of Hami melon by the technology of hyperspectral imaging was put forward. The research used the hyperspectral imaging system to get the diffuse reflective spectrum information (400～1 000 nm) of anilox class Hami melon sugar content, chose effective whole wavelength(500～820 nm)to do the modeling regression analysis the sugar content of Hami melon. The research compared the correction method of MSC and SNV, and also compared the influence of accuracy of modeling in terms of the spectrum pretreatment methods of original spectrum, first order differential, second order differential; Using the methods of PLS, SMLR and PCR, the comparative analysis of sugar content detection model effect with skin Hami melon and peel Hami melon was conducted. The results showed that after the original spectrum being processed by MSC and first order differential spectrum, modeling effect could be very good using the method of PLS and SMLR. Synthesizing correction set correlation coefficient and forecast modeling effect, it’s feasible to detect the sugar content of skin Hami melon by the PLS method, with a correction sample correlation coefficient (R_c) of 0.861 and the lower root mean square errors of correction (RMSEC) of 0.627, and a prediction sample correlation coefficient (R_p) of 0.706 and root mean square errors of prediction (RMSEP) of 0.873. The best effect to detecti the sugar content of peel Hami melon was obtained by the SMLR method with a correction sample correlation coefficient (R_c) of 0.928 and the lower root mean square errors of correction (RMSEC) of 0.458, with a Prediction sample correlation coefficient (R_p) of 0.818 and root mean square errors of prediction (RMSEP) of 0.727. The results of this study indicate that the technology of hyperspectral imaging can be used to predict the sugar content of Hami melon.