Profile-based diffuse reflectance corrections for improved optical property measurement of spherical fruit with spatial frequency domain imaging

Diffuse reflectance (Rd), demodulated from spatial frequency domain imaging (SFDI), is strongly influenced by the surface profile of target sample, and the optical properties (absorption coefficient μa and reduced scattering coefficient μ's) extracted from the Rd are prone to large estimation errors caused by curved surface of the fruit. This study proposes a profile-based Rd correction method, aiming to improve the accuracy of optical property measurement for spherical fruit, such as apples, as well as enhance fruit bruise detection through improved optical properties. Firstly, the surface height of the test object was obtained by performing the phase-measuring profilometry (PMP) in SFDI. Secondly, the object surface reflection angle was calculated using the proposed angle calculation method for spherical samples. The Modified Lambertian Correction (MLC) method was then refined to minimize the effect of the incident angle on the Rd correction. The performance of the Rd correction method was validated using standard reflector and Teflon spheres. The results showed that the average error of the PMP measurement with the height range of 0–50 mm was within 1 mm, the relative error of the corrected Rd was within 2 %, and the standard deviation of the Rd for the Teflon sphere was reduced by about 60 % after correction. Finally, the proposed Rd correction method was applied to measure the optical properties of four spherical fruit ('Golden Delicious', 'Red Fuji' apples, lemons, and oranges). The results revealed that the standard deviations of the measured μa and μ's of the spherical fruit were greatly reduced by about 70 % and 40 % after correction. The improved optical property mappings enhanced the early bruise detection of apples due to the fact that the optical property contrast between the non-bruised and bruised tissues were significantly increased.