Moisture measurement of tea leaves during withering using multifrequency microwave signals optimized by ant colony optimization

Abstract

The rapid and accurate moisture content (MC) measuring of withering tea leaves faces multiple challenges. In this study, a self-developed microwave sensing system equipped with multifrequency-swept signals (2.00–10.00 GHz, an interval of 10.00 MHz) was used to measure the MC of withering tea leaves (16.25%–77.65%). Four kinds of hybrid-based ant colony optimizations were advanced to optimize microwave frequency signals, and the attenuation features at six frequencies (2.16, 3.42, 3.88, 4.48, 5.32 and 7.70 GHz) and phase shift features at four frequencies (2.84, 4.04, 7.88 and 8.43 GHz) were selected. Based on those optimized features, three primary models (support vector regression, extreme gradient boosting, and multi-layer perception) and one secondary model (decision tree) presented the best MC values (the average of mean absolute error was 0.349). The results showed the potential application of microwave technology in the MC monitoring in tea processing industry. • A microwave sensing system was developed for moisture detection in tea withering. • A hybrid-based ACO algorithm was designed to optimize the frequency features. • Attenuation and phase shift features correlated with MC were selected by MRACO. • The best MC detection was achieved by the stacking model of SVR, LGBM, and MLP.