Development of a multispectral imaging prototype for real-time detection of apple fruit firmness

Abstract

Multispectral scattering is a promising nondestructive technique for assessing the firmness of fruit. We report on the development of a laser-based multispectral imaging prototype for real-time detection of apple firmness. The prototype consists of a common aperture multispectral imaging unit, a specially designed multilaser unit, and a fruit handling unit, which simultaneously capture and process four spectral scattering images at a speed up to two fruit per second. The multispectral imaging system is tested for detecting the firmness of Golden Delicious and Red Delicious apples at a rate of one fruit every two seconds. The original 2-D scattering images are corrected by removing noise pixels and incorporating fruit size into the calculations of the scattering distance and intensity. The 1-D scattering profiles are fitted with a four-parameter Lorentzian distribution function. Multilinear regression models are developed using the four Lorentzian parameters for the four wavelengths to predict the firmness of apples. The multispectral imaging system achieves good firmness predictions with the correlation coefficient of 0.86 for both Golden Delicious and Red Delicious. The technique is fast and relatively easy to implement, and has the potential to meet the requirement for online sorting and grading of apples.