基于小波变换的水果糖度近红外光谱检测研究

利用小波变换滤波技术对90个水果样品的近红外光谱信号进行了去噪处理,并结合滤波后重构光谱信号对水果糖度进行逐步线性回归(SMLR)建立其校正模型,通过34个样品的外部检验对校正模型精度进行评价.研究结果表明:校正模型的预测精度在小波尺度为3时其预测精度最好,预测集的决定系数由原来的0.84提高到0.85,预测集相对标准误差由原来的6.1%降为6.0%.因此,使用小波去噪方法有消除原始光谱噪声作用,从而使最终的SMLR模型更具有代表性和稳健性,也提高了品质检测时模型预测精度.

Sugar Content Prediction of Apple Using Near-Infrared Spectroscopy Treated by Wavelet Transform

Based on wavelet transform (WT) by using the difference in wavelet modulus maxima evolution behaviors between singular signals and random noises in multi-scale space, the near infrared spectroscopic signals of 90 fruit samples were denoised by wavelet transform. The sugar content in intact apple was calculated by stepwise regression method. The result of calibration model after noise filtering was satisfactory. The relative standard error of prediction is reduced to 6.0% from 6.1% of original spectra. It is concluded that wavelet transform is an useful method to eliminate noise of NIR signals, as it makes the final calibration model more representative and stable and robust.