Detection Of Soluble Solids Content In Specialty Citrus Of Jiangxi By VIS/NIR Spectroscopy

Specialty citrus of Jiangxi (Nanfeng mandarin and Gannan navel orange) was usedas the research object, and the visible/near-infrared (Vis/NIR) spectroscopy andchemometrics methods were used to detect soluble solids content (SSC) of Nanfengmandarin and Gannan navel orange. Some influence factors that may affect theperformance of SSC model were investigated, and a series of pre-treatment methods, avariety of variable selection methods as well as four kinds of modeling methods wereused to optimize the SSC model. The main results and conclusions are listed as follows:1) The effect of two spectral detection modes on SSC model performance of Nanfengmandarin and Gannan navel orange was studied. The results show that the effect ofspectral detection modes on the performance of SSC model is significant,transmission detection mode is better than reflection detection mode. Moreover, theSSC model established by the average of three spectra is superior to single spectrum.2) Based on transmission detection mode, the effect of different orientation on theperformance of SSC model of Gannan navel orange was studied. The results showthat when the angel composed by navel orange stems and pit as well as incident lightis vertical (90°), the prediction models of SSC in Gannan navel orange is the best,the random (not including0°and90°) is next, and the parallel (0°) is the worst.3) Based on transmission detection mode and the optimum orientation, the effect of peelon SSC model of Nanfeng mandarin and Gannan navel orange was studied. Theresult shows that the effect of peel on the performance of SSC model is remarkable4) SSC models of Nanfeng mandarin and Gannan navel orange were optimized using avariety of variable selection methods combined with four calibration methods. First,PLS with a variety of variable selection methods were used to develop SSC models,and it was found that competitive adaptive reweighted sampling (CARS) was thebest variable selection method. Then in order to find the best calibration method,CARS together with another four kinds of calibration methods were used to establishSSC models. The result shows that CARS combined with principal componentregression (PCR) is the best combination for Nanfeng mandarin, and CARS-PLS isthe best combination for Gannan navel orange.