基于小波变换-遗传算法-偏最小二乘的草莓糖度检测研究

Detection of Sugar Degree in Strawberry Based on Wavelet Transform-Genetic Algorithm-Partial Least Squares

采用近红外漫反射光谱分析技术,对草莓糖度进行了无损检测研究。利用便携式近红外光谱仪采集草莓样品在600~1100 nm波段内的漫反射光谱数据。首先利用小波变换(WT)多分辨率方法对光谱数据进行去噪预处理,然后利用遗传算法(GA)优选特征波长,最后运用偏最小二乘法(PLS)建立草莓糖度的WT-GA-PLS校正模型。该模型校正集的相关系数R C为0.9395,校正集的均方根误差RMSEC为0.1615,预测集的相关系数R P为0.9652,预测集的均方根误差EMSEP为0.5042。与全光谱模型(FS-PLS)和小波变换模型(WT-PLS)相比,该模型预测能力更强,稳健性更优。

Near infrared(NIR)diffuse reflectance spectroscopy was used to study the non-destructive detection of strawberry sugar content.Firstly,the diffuse reflectance spectra of strawberry samples in 600-1100 nm band were collected by portable near infrared spectrometer.The wavelet transform(WT)multi-resolution analysis method was used to denoise the original spectral data,and the spectral data with obvious characteristic peaks and clear contour were obtained.The genetic algorithm(GA)algorithm was used to optimize the selection of 2001 wavelength points in 600-1100 nm band,and 201 wavelength points with a cumulative contribution of more than 50%were obtained.Partial least squares regression(PLS)was used to establish the WT-GA-PLS calibration model of strawberry.The correlation coefficient RC of the calibration set is 0.9395,the RMSEC of the calibration set is 0.1615,the correlation coefficient RP of the prediction set is 0.9652,and the EMSEP of the prediction set is 0.5042.Compared with the full spectrum model(FS-PLS)and the wavelet transform model(WT-PLS),the model has better prediction ability and robustness.The application of portable spectral analysis technology provides a theoretical basis for the non-destructive detection of strawberry sugar content,also provides a feasibility for long-term monitoring of fruit dynamic changes,the realization of orchard management and post-harvest detection.