Moisture content assessment of dried Hami jujube using image colour analysis

Abstract

To investigate the feasibility of image colour information in predicting the moisture content of dried Hami jujube, the images were obtained under different colour space models, and the colour model component mean and chromaticity frequency sequences of R, G, B, H, S, V, L*, a* and b* were extracted through image analysis. After optimising the colour model component mean and chromaticity frequency sequence, the model was established and compared. The results showed that the GA-ELM (genetic algorithm - extreme learning machine) model established by CARS (competitive adaptive reweighted sampling) method to optimise 12 chromaticity features of S chromaticity frequency sequence had the best prediction effect, with R<sub>c</sub> of 0.917, R<sub>p</sub> of 0.934 and residual predictive deviation (RPD) of 2.507. Therefore, the colour image information can accurately predict the moisture content of dried Hami jujube.