Abstract:
A low-cost sensor is developed for non-destructive fruit quality estimation which switches
light-emitting diodes sequentially on at 4 various wavelengths and measures the reflection in the
interactive mode. The detector is tested on Apple samples for a non-destructive estimation of
soluble sugar content (SSC), and dry matter (DM). A total of 240 apple samples were determined
for SSC and DM measurements. The developed handheld device is composed of a Raspberry Pi,
analog to digital converter ADS1115, light emitting diodes (LEDs), a liquid crystal display (LCD)
screen, a photodetector, resistors, PCBs, buttons, headers, jumper wires, batteries and 18650
batteries shield. The photodetector was used for linear regression i.e., partial least square (PLS)
regression and multiple linear regression (MLR) and support vector machine (SVM) as non-linear
regression while the benchmark spectrometer i.e., Felix F-750, was fitted with a partial least square
(PLS) regression model. The best estimate for the two different measurements were employed in
different regression techniques. A LED detector was observed for SSC measurements with
correlation coefficient (R) of 0.82 and 1.51% of root mean square error for prediction (RMSEP).
whereas SSC values were 0.95 and 0.76%, respectively with the benchmark spectrometer. The
LED detector prototype achieved DM values of R = 0.85 and RMSEP =1.59%, while for
benchmark spectrometer, R and RMSEP of DM was 0.93 and 0.69%, respectively.
