A Wavelength Selection Model Based on Successive Projections Algorithm for pH Detection of Water by VIS-NIR Spectroscopy

Abstract A novel pH detection method using VIS-NIR (visible near infrared) spectroscopy for water is introduced in this paper. In order to improve prediction accuracy and calculation speed of pH detection using VIS-NIR spectroscopy, a wavelength selection model of SPA-MLR (successive projections algorithm and multiple linear regression) is proposed. Two experiments (pH measurement experiment by VIS-NIR spectrophotometer and pH measurement experiment by VIS-NIR grating spectrograph) are employed to evaluate the performance of SPA-MLR. At meanwhile, PLS (partial least squares), GA-PLS (genetic algorithm and partial least squares) are introduced to compare with SPA-MLR. The results indicate that, the prediction accuracy and calculation speed of SPA-MLR are all better than that of PLS and GA-PLS. Under the pH measurement experiment using VIS-NIR spectrophotometer, only one wavelength is selected from 351 original wavelengths for calibration, the RMSEP (root mean square errors of prediction) is 0.35, and the model training speed is 2.36s. Under the pH measurement experiment using VIS-NIR grating spectrograph, 3 wavelengths are selected from 2860 original wavelengths for calibration, the RMSEP is 0.50, and the model training speed is 56.45s. For the two experiments, the selected wavelength variables by SPA are distributed mainly in the band of 870 nm to 990 nm.