Banana Maturity Characteristic Prediction Based on Hyperspectral and PCA-ELM
Article
Figures
Metrics
Preview PDF
Reference
Related
Cited by
Materials
Abstract:
Spectral data of 450 unbarked banana samples, stored at room temperature, were collected by hyperspectral imaging tecchique. The single factor analysis of variance was used to measure the soluble solid content (TSS) and firmness (FIM), and then a spectral and image feature classification model of banana maturity physicochemical index was established using partial least squares linear regression analysis method of ridge optimization (RR-iPLS). The results showed that the correlation coefficient values (R2) of the soluble solid content of spectral data predicting banana and the firmness were 0.92 and 0.94. Then, characteristic wavelengths were selected by continuous projection method (successive projections algorithm, SPA) and principal component analysis (principal component, analysis, PCA) and the extreme learning machine (extreme learning machine, ELM) was established for the cross validation of spectral data based on characteristic wavelengths. , The PCA-ELM classification model based on characteristic wavelength had a better prediction performance with a high accuracy by comparing the RR-iPLS, SPA-ELM and PCA-ELM classification prediction models. The accuracy of cross validation reached to 99%, which could provide an effectively predictive study for the rapid and non-destructive identification of banana quality. The proposed method basically fulfilled the classification and detection of banana maturity and could achieve effective economic benefits.
