Product Quality Estimation of Sago (Metroxylon sagu) Based on Hyperspectral Imaging and Multivariate Image Analysis

Abstract

The quality monitoring process for sago often relies on traditional lab instruments, seen as complex, expensive, and time-consuming. To tackle such issues, this project, therefore, aims to develop an efficient sago quality estimator based on hyperspectral imaging (HSI) with multivariate analysis. The newly proposed Adaptive 1D-ConvNet architecture developed one of the best-performing models, achieving Rp2 of 0.9410 to 0.9981 and RPD of 4.11 to 32.08. In conclusion, the HSI combined with multivariate analysis proved effective as a rapid, reliable, and cost-effective sago quality estimator.