Detecting Surface Defects of Achacha Fruit (Garcinia humilis) with Hyperspectral Images

Hyperspectral imaging data within the wavelength range of 400–1000 nm were used to classify common skin conditions (i.e., normal, scar, decay, and insect bite) achacha fruits. The band ratio (BR) spectral angle mapper (SAM) algorithms in a binary classification. Furthermore, SAM, support vector machine (SVM), artificial neural network (ANN) models multiclass performances classification assessed. For binary-classification approach, three defective classes merged into one, accuracies BR (990 nm/600 nm) SAM 78.70% 75.02%, respectively. SVM, ANN four class problems 58.36%, 83.59%, 99.88%, A principal component analysis (PCA) was for reduction. Nine characteristic wavelengths extracted from weighting-coefficient curves first components. Using only nine selected bands, 51.49%, 80.76%, 96.85%, Compared with using full bands decreased slightly; however, gain speed potential data-acquisition can expedite