Accurate Ripening Stage Classification of Pineapple Based on a Visible and Near-Infrared Hyperspectral Imaging System

Area of Science:

• Agricultural Science
• Food Science
• Spectroscopy

Background:

• Pineapples are economically significant tropical fruits.
• Assessing pineapple ripeness is critical for harvesting, marketing, and processing.
• Ensuring optimal pineapple quality requires accurate ripeness determination.

Purpose of the Study:

• To determine the optimal classification model for pineapple ripening stages.
• To analyze spectral information and soluble solid content (SSC) during pineapple ripening.
• To utilize hyperspectral imaging (HSI) for non-destructive ripeness assessment.

Main Methods:

• Employed visible and near-infrared hyperspectral imaging (VIS-NIR-HSI) from 400-1000 nm.
• Applied four preprocessing methods (SNV, MSC, normalization, SG smoothing).
• Utilized feature extraction algorithms (SPA, BOSS) with machine learning classifiers (SVM, ELM, KNN, RF).

Main Results:

• Random Forest (RF) with SNV preprocessing achieved 94.44% accuracy.
• A total of 33 wavelengths selected by BOSS combined with RF yielded a 97.22% test accuracy.
• The study identified optimal wavelengths for ripeness classification.

Conclusions:

• Near-infrared hyperspectral imaging (NIR-HSI) is a viable non-destructive tool for pineapple ripeness identification.
• This method can enhance the classification and grading of pineapples, addressing quality issues.
• Developing low-cost multispectral systems based on key wavelengths is feasible for industrial applications.