可溶性固形物含量(SSC)和硬度是哈密瓜划分等级的重要指标, 同时也是其成熟度的表征因子。 因此, 为满足哈密瓜自动化分级和适宜采摘, 采用高光谱技术结合特征波长筛选的方法, 同时对哈密瓜的可溶性固形物含量、 硬度及成熟度进行了无损检测研究。 对多元散射校正(MSC)处理后的光谱分别利用连续投影算法(SPA)、 竞争性自适应重加权算法(CARS)和CARS-SPA方法筛选了哈密瓜可溶性固形物和硬度的特征波长, 并将原始光谱、 MSC预处理后的光谱和所筛选的特征波长作为输入变量分别建立哈密瓜可溶性固形物和硬度的支持向量机(SVM)预测模型及成熟度判别模型。 结果显示, MSC-CARS-SPA方法所建立的可溶性固形物和硬度SVM预测模型最优, 其Rpre, RMSEP和RPD分别为0940 4, 0402 7, 294 1和0825 3, 3522, 1771。 同时对哈密瓜成熟度进行了判别分析, 并分别建立了基于全光谱、 单一的可溶性固形物或硬度特征波长和主成分分析(PCA)特征融合的哈密瓜成熟度SVM判别模型。 结果显示, CARS-PCA-SVM模型的判别结果与全光谱SNV-SVM模型相同, 其校正集和预测集判别正确率分别为95%和94%。 研究表明, 利用高光谱技术结合特征波长筛选方法可实现同时对哈密瓜可溶性固形物和硬度的定量预测及成熟度判别。

Soluble solids content (SSC) and firmness are not only important indicators for grading of Hami melon but also characteristic factors to determine its maturity. Thus, in order to achieve automatic grading and suitable picking of Hami melon, hyperspectral imaging technology combined with different characteristic wavelengths selection methods were used to simultaneously assess SSC, firmness and maturity of Hami melon. Successive projections algorithm (SPA), competitive adaptive reweighted sampling (CARS) and CARS-SPA algorithm were used to select the characteristic wavelengths of SSC and firmness of Hami melon from MSC pretreated spectra. The full spectral variables and selected wavelength variables were used as the inputs to build SVM model for determination of the SSC, firmness and maturity of Hami melon, respectively. The results indicated that the MSC-CARS-SPA-SVM models achieved the optimal performance for SSC and firmness of Hami melon. The correlation coefficient of prediction set (Rpre) , the root mean square error of prediction (RMSEP ) and the relative prediction deviation (RPD) were 0940 4, 0402 7 and 2941 for SSC and 0825 3, 3522 and 1771 for firmness, respectively. At the same time, the full spectrum, selected characteristic wavelengths for SSC or firmness and feature fusion by the principal component analysis (PCA) were used to build SVM discriminatory models for maturity of Hami melon, respectively. The results showed that the discriminant results of CARS-PCA-SVM model was agreement with the FS-SNV-SVM model, the recognition rate of calibration set and prediction set were 95% and 94%. The research indicated that it is the feasible to use hyperspectral imaging technology combined with different characteristic wavelengths selection methods can be used to evaluate SSC, firmness and maturity of Hami melon simultaneously.