Submitted to: Proceedings of SPIE
Publication Type: Proceedings
Publication Acceptance Date: 12/1/2003
Publication Date: 1/1/2004
Citation: Lu, R. 2004. Near infrared multispectral scattering for assessing internal quality of apple fruits. Proceedings of SPIE. 5271:313-320.
Interpretive Summary: Firmness and sweetness are key attributes that determine the acceptability of fresh apple fruit to the consumer. Technologies that can grade and sort fruit for firmness and sweetness would greatly improve the U.S. fruit industry's ability to compete in the global marketplace and better meet consumer's increasing demands for fruit quality. Our objective was to investigate a multispectral imaging system, which can acquire spectral images in four discrete wavelengths simultaneously, for predicting fruit firmness and soluble solids content (or total sugar). The multispectral imaging system was assembled and tested with Red Delicious and Golden Delicious apples. Computer algorithms were developed to analyze light scattering characteristics and relate them to fruit firmness and soluble solids content. The multispectral imaging system predicted fruit firmness and soluble solids content with the correlation coefficient (r) of 0.76 and 0.77, respectively, for Red Delicious and r=0.73 and 0.82, respectively, for Golden Delicious. This research demonstrated that the multispectral imaging system is capable of real time sensing of fruit internal quality. With further improvement, the technique can be used for sorting and grading fruit for firmness and sweetness. This would provide the fruit industry with a means to deliver superior quality and consistent fresh products to the marketplace.
Technical Abstract: Firmness and sweetness are key quality attributes that determine the acceptability of apple fruit to the consumer. The objective of this research was to investigate a multispectral imaging system for simultaneous acquisition of multispectral scattering images from apple fruit to predict firmness and soluble solids content (SSC). A circular broadband light beam was used to generate light backscattering at the surface of apple fruit and scattering images were acquired, using a common aperture multispectral imaging system, from Red Delicious and Golden Delicious apple fruit for wavelengths at 680, 880, 905, and 940 nm. Scattering images were radially averaged to produce one-dimensional spectral scattering profiles, which were then entered into a backpropagation neural network for predicting apple fruit firmness and SSC. Our results showed that the neural network performed best when 10 neurons and 20 epochs were used. With inputting three ratios of spectral profiles involving all four wavelengths, the neural network gave firmness predictions with the correlation (r) of 0.76 and the standard error for validation (SEV) of 6.2 N for Red Delicious apples and r=0.73 and SEV=8.9 N for Golden Delicious apples. Relatively good SSC predictions were obtained for both varieties with SEV=0.9 deg Brix.