|Peng, Yankun - MICHIGAN ST UNIVERSITY|
Submitted to: Fruit, Nut, and Vegetable Production Engineering International Symposium
Publication Type: Abstract Only
Publication Acceptance Date: December 15, 2004
Publication Date: September 12, 2005
Citation: Lu, R., Peng, Y. 2005. Comparison of multispectral scattering and visible/NIR spectroscopy for predicting apple fruit firmness [abstract]. Fruit, Nut, and Vegetable Production Engineering International Symposium. Technical Abstract: Considerable recent research has been reported on using visible/near-infrared (visible/NIR) spectroscopy for measuring internal quality of fruits. However, reported results showed the difficulty of using visible/NIR spectroscopy to predict fruit firmness accurately. We recently proposed a new method of using multispectral/hyperspectral scattering for predicting apple fruit firmness and relatively good results were obtained. The overall objective of this research was to further compare multispectral scattering with visible/NIR spectroscopy for predicting fruit firmness. Specific objectives were to: 1) develop a multispectral imaging system that can acquire scattering images from apple fruit for wavelengths in the visible and short-wave NIR region; 2) propose an improved mathematical model for describing spectral scattering profiles to predict fruit firmness; and 3) compare the multispectral scattering technique with visible/NIR spectroscopy for predicting apple fruit firmness. A liquid crystal tunable filter based multispectral imaging system with a low cost CCD camera was developed for measuring spectral scattering from apple fruit for wavelengths between 650 nm and 1,000 nm. For comparison, a visible/NIR spectroscopy system was assembled for acquiring diffuse reflectance spectra from apple fruit in interactance mode over the wavelengths between 500 nm and 1,100 nm. Experiments were performed with the two systems to acquire multispectral scattering images and reflectance spectra from Red Delicious and Golden Delicious apples. Fruit firmness was measured from individual apples with the standard destructive Magness-Taylor (MT) firmness testing technique. A Lorentzian distribution function with four parameters was proposed for describing multispectral scattering profiles from apple fruit. Spectra of Lorentzian parameters were analyzed for determining an optimal set of wavelengths for predicting fruit firmness. A multi-linear regression model was developed for relating Lorentzian parameters to fruit firmness. Standard chemometric procedures were used to analyze visible/NIR spectral data; a calibration model was developed by using the partial least squares method. The Lorentzian distribution function gave excellent fits to the spectral scattering profiles, including both saturation and scattering areas, with the average r equal to 0.999. Best firmness predictions from the multispectral scattering system were obtained with seven wavelengths (690, 770, 790, 810, 920, 980, and 1000 nm) for Red Delicious and eight wavelengths (650, 690, 740, 750, 820, 880, 910, and 990 nm) for Golden Delicious. With the optimal wavelengths, the multispectral scattering system predicted MT firmness with r=0.82 and 0.81 for Red Delicious and Golden Delicious, respectively, versus r=0.50 and 0.48 from visible/NIR spectroscopy. This research demonstrated that multispectral scattering is superior to visible/NIR spectroscopy for predicting apple fruit firmness. The multispectral scattering technique is potentially useful for sorting and grading apples and other fruits for firmness.