Quantitative prediction and visual detection of the moisture content of withering leaves in black tea (Camellia sinensis) with hyperspectral image

In actual production, the rapid and non-destructive detection of moisture content in withering leaves is still an unsolved problem due to the randomness of withering leaves on the conveyor belt and the limitation of detection range using near infrared detection equipment. To solve this problem, the hyperspectral images of accumulated withering leaves were obtained and the moisture prediction model was established in the range of 400–1000 nm. Different pretreatment and effective bands selection methods were used to optimize the model. The results showed that the performance of nonlinear model was better than that of linear model and the SNV-Si-CARS-ELM model had the best performance. The Rc2, RMSEC, Rp2, RMSEP and RPD were 0.9940, 0.0074, 0.9942, 0.0078 and 13.0907, respectively. Furthermore, the moisture distribution maps of accumulated withering leaves in different withering degrees were developed. This study provides a theoretical basis for the on-line application of hyperspectral image technology in black tea processing.