Hyperspectral and Deep Learning-based Regression Model to Estimate Moisture Content in Sea Cucumbers

Abstract

The hyperspectral image technology contains information in spectral and spatial forms that produce a huge amount of data. This data becomes an additional load while data is processed. Deep learning is the latest method capable of processing large-scale data with a deep structure of artificial neural network (ANN) and improving the model performance of data analysis. Therefore, this study aims to get a deep learning model into hyperspectral image processing for quantitative measurements of moisture content in dried sea cucumbers study case. The sea cucumber used in this study is the dried sea cucumber (Holothuria scabra), commonly known as Beche-de-mer. This study used the 400–1000 nm wavelength range to measure the moisture content quickly and nondestructively. The proposed model is deep learning which is used to build a predictive model system for moisture content in dried sea cucumbers. The coefficient of determination and the root means square error evaluate the measurement system. The measurement results of moisture content, the coefficient of determination, and the root mean square error values for training data are 0.99 and 0.11%, while testing data are 0.92 and 0.29%.