کد مقاله | کد نشریه | سال انتشار | مقاله انگلیسی | نسخه تمام متن |
---|---|---|---|---|
4517756 | 1624975 | 2016 | 11 صفحه PDF | دانلود رایگان |
![عکس صفحه اول مقاله: Multispectral scattering imaging and NIR interactance for apple firmness predictions Multispectral scattering imaging and NIR interactance for apple firmness predictions](/preview/png/4517756.png)
• Multispectral scattering imaging was inferior to NIR interactance method.
• Multiple linear regression significantly improved the correlation to firmness.
• Modified Lorenzian model outperformed diffusion model.
• Modified Lorenzian model showed similar results on intact and cut apples.
A multispectral imaging (MSI) system, using four discrete wavelengths (685, 850, 904 and 980 nm), has been developed and validated for making spatially resolved reflectance spectroscopy (SRRS) measurements. The primary aim was to evaluate the potential of MSI for high-speed firmness grading of apples. The MSI system validations were made using Intralipid solutions of known concentration and comparing the results against measurements made using a laboratory based inverse adding-doubling method (IAD). The results compared well for scattering properties with both the MSI and IAD measurements in reasonable agreement with known properties. For the absorption properties only the MSI measurements were close. Performance of the MSI system was then compared with a near-infrared spectroscopy (NIRS) system using 100 ‘Royal Gala’ apples (Malus domestica Borkh.). The apples were measured non-destructively by both the MSI and NIRS systems. Cut apple surfaces were also examined by the MSI system and excised slices of apple tissue were measured using the IAD system. Actual apple firmness was measured by destructive penetrometer. The MSI data analysis involved use of both a phenomenological diffusion model and a heuristic modified Lorentzian model for describing the scattering images at each wavelength. The best MSI results of R = 0.87 and RMSECV = 7.17 N were obtained when the Lorentzian model parameters derived at each wavelength were combined using multiple linear regression (MLR). The NIRS system measurements were still a little better, with a best correlation on R = 0.90 and RMSECV = 6.99 N.
Journal: Postharvest Biology and Technology - Volume 119, September 2016, Pages 58–68