NIR Hyperspectral imaging for Hardness of Gari/Eba (cooked dough), at IITA, Ibadan, Nigeria

The near-infrared hyperspectral images of freshly prepared Gari_Eba produced from 23 cassava genotypes were acquired using the SPECIM FX 17 Camera. Gari_Eba dough was homogenized and placed on the translation stage for a line scan image capture. The hyperspectral image was captured in the 900 to 1700 nm wavelength range and has a spectral resolution of 8 nm. Reflectance calibration and image segmentation were carried out using the Breeze Software (version 2024.0.1) to extract the average spectra from each image. An attempt to predict the textural attributes of Gari_Eba was performed by combining the samples’ average spectra with the reference texture values. NIR hyperspectral imaging demonstrated a promising prediction performance for the hardness of Eba but could not accurately predict other textural attributes of Eba. Further attempts will be made by extracting the effective wavelengths from the average raw spectra to improve the model's accuracy by removing the redundant information. The coefficient of determination in calibration for hardness was 0.90 and prediction performance of 0.76 respectively.