Estimating blueberry mechanical properties based on random frog selected hyperspectral data

• A hyperspectral reflectance and transmittance imaging setup was developed.
• Blueberry mechanical properties were predicted by random frog algorithm.
• Hyperspectral data simultaneously predicted several mechanical parameters well.
• Combined spectra improved prediction performance for some mechanical parameters.

A hyperspectral reflectance and transmittance imaging system was developed to non-destructively evaluate the comprehensive mechanical properties of blueberry. Reflectance and transmittance spectra were extracted from segmented hyperspectral images of whole fruit and correlated with fruit mechanical properties obtained from texture profile analysis and puncture analysis using least squares-support vector machine. A random frog spectral selection approach was applied to collect informative wavelengths. Prediction models based on random frog selected reflectance and transmittance spectra gave similar results to those based on respective full spectra. Combined spectra with single random frog, which were obtained by combining random frog selected reflectance and transmittance into one spectral vector, were feasible for predicting hardness, springiness, resilience, force max and final force, with Rp (RPD) values of 0.86 (1.78), 0.72 (1.73), 0.79 (1.78), 0.77 (1.51) and 0.84 (1.72), respectively. When applying random frog again for combined spectra with single random frog, the obtained models were also satisfactory with fewer wavelengths. In conclusion, the use of hyperspectral reflectance and transmittance as well as their combined spectra, coupled with random frog approach, showed a considerable potential for predicting blueberry mechanical properties.