为探讨小波压缩算法结合近红外光谱技术在马铃薯全粉还原糖含量检测中的可行性, 采用傅里叶变换近红外光谱仪采集了250份马铃薯全粉样品的近红外光谱。 分别优化了消失矩、 小波系数和主成分因子数, 优化结果为10, 100和20。 基于db小波函数将1 501个马铃薯全粉的近红外光谱变量压缩成100个小波系数。 分别以1 501个光谱变量和100个小波系数为变量分别建立了偏最小二乘(PLS)校正模型。 以62个未参与建模的样品作为预测集, 考察模型的预测能力。 经比较, 小波压缩结合PLS的校正模型预测结果最优, 模型预测相关系数为0.98, 预测均方根误差为0.181%。 实验结果表明小波压缩算法结合近红外光谱技术有效地保留了有效光谱信息, 实现了光谱数据降维, 简化了马铃薯全粉还原糖PLS校正模型, 提高了模型的预测能力。

The feasibility was explored in determination of reducing sugar content of potato granules based on wavelet compression algorithm combined with near-infrared spectroscopy. The spectra of 250 potato granules samples were recorded by Fourier transform near-infrared spectrometer in the range of 4 000～10 000 cm-1. The three parameters of vanishing moments, wavelet coefficients and principal component factor were optimized. The optimization results of three parameters were 10, 100 and 20, respectively. The original spectra of 1 501 spectral variables were transfered to 100 wavelet coefficients using db wavelet function. The partial least squares (PLS) calibration models were developed by 1 501 spectral variables and 100 wavelet coefficients. Sixty two unknown samples of prediction set were applied to evaluate the performance of PLS models. By comparison, the optimal result was obtained by wavelet compression combined with PLS calibration model. The correlation coefficient of prediction and root mean square error of prediction were 0.98 and 0.181%, respectively. Experimental results show that the dimensions of spectral data were reduced, scarcely losing effective information by wavelet compression algorithm combined with near-infrared spectroscopy technology in determination of reducing sugar in potato granules. The PLS model is simplified, and the predictive ability is improved.