Skip to main content
ARS Home » Midwest Area » East Lansing, Michigan » Sugarbeet and Bean Research » Research » Publications at this Location » Publication #173959


item Lu, Renfu

Submitted to: Biosystems Engineering
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/3/2005
Publication Date: 1/4/2006
Citation: Lu, R., Peng, Y. 2006. Hyperspectral scattering for assessing peach fruit firmness. Biosystems Engineering. 93(2):161-171.

Interpretive Summary: Currently, fruit processors and packers are relying on automatic sorting and handling technologies to achieve high production efficiency and meet market demands for fruit quality. Machines that are capable of sorting fruit for size, weight and color have been developed and used commercially. However, consumers are less concerned about fruit appearance than quality attributes such as firmness and taste. Techniques that can sort and grade for these important quality attributes are essential to delivering high quality, consistent fresh products to the marketplace and meeting increased consumer expectations for fruit quality. Our research proposed a new method of using light scattering at multiple wavelengths to estimate fruit firmness. A hyperspectral imaging technique was used to measure light scattering from peaches over the visible and near-infrared (long than the visible) wavelengths. Hyperspectral imaging is an advanced imaging technique that provides spectral information for individual image pixels of an object. A mathematical model coupled with artificial neural networks was proposed for relating spectral scattering features to peach fruit firmness. Good firmness predictions were obtained with a correlation coefficient of 0.90. The hyperspectral scattering technique is nondestructive and rapid, and it is promising for assessing, sorting and grading peaches for firmness.

Technical Abstract: Firmness is an important quality parameter in determining optimal harvest time and appropriate postharvest handling and marketing strategies. The objective of this research was to investigate the potential of using hyperspectral scattering to predict peach fruit firmness. A hyperspectral imaging system was used to acquire spectral scattering images from 'Red Haven' peaches over the spectral region between 500 nm and 1040 nm. A Lorentzian distribution function with two parameters was proposed to describe the spectral scattering profiles of peach fruit for individual wavebands. Principal component analysis (PCA) was performed on spectra of Lorentzian parameters and their product to extract essential spectral scattering information and reduce data sizes. PC scores from parameter spectra were input into a backpropagation feedforward neural network for predicting fruit firmness. Among the three spectra, the product of Lorentzian parameters gave best firmness predictions with r=0.90 and the standard error for validation of 12.07 N. Hyperspectral scattering is useful for assessing the firmness of peach fruit.