Nondestructive Determining the Soluble Solids Content of Citrus Using Near Infrared Transmittance Technology Combined with the Variable Selection Algorithm
Xi Tian,Jiangbo Li,Shilai Yi,Guoqiang Jin,Xiaoying Qiu,Yongjie Li
DOI: https://doi.org/10.1016/j.aiia.2020.05.001
2020-01-01
Artificial Intelligence in Agriculture
Abstract:Nondestructive determination the internal quality of thick-skin fruits has always been a challenge. In order to investigate the prediction ability of full transmittance mode on the soluble solid content (SSC) in thick-skin fruits, the full transmittance spectra of citrus were collected using a visible/near infrared (Vis/NIR) portable spectrograph (550–1100 nm). Three obvious absorption peaks were found at 710, 810 and 915 nm in the original spectra curve. Four spectral preprocessing methods including Smoothing, multiplicative scatter correction (MSC), standard normal variate (SNV) and first derivative were employed to improve the quality of the original spectra. Subsequently, the effective wavelengths of SSC were selected from the original and pretreated spectra with the algorithms of successive projections algorithm (SPA), competitive adaptive reweighted sampling (CARS) and genetic algorithm (GA). Finally, the prediction models of SSC were established based on the full wavelengths and effective wavelengths. Results showed that SPA performed the best performance on eliminating the useless information variable and optimizing the number of effective variables. The optimal prediction model was established based on 10 characteristic variables selected from the spectra pretreated by SNV with the algorithm of SPA, with the correlation coefficient, root mean square error, and residual predictive deviation for prediction set being 0.9165, 0.5684°Brix and 2.5120, respectively. Overall, the full transmittance mode was feasible to predict the internal quality of thick-skin fruits, like citrus. Additionally, the combination of spectral preprocessing with a variable selection algorithm was effective for developing the reliable prediction model. The conclusions of this study also provide an alternative method for fast and real-time detection of the internal quality of thick-skin fruits using Vis/NIR spectroscopy.