AI-Based Expert System for Sugar Apples Fruit Quality and Ripeness Assessment

Authors: Kuo-Dung Chiou, Yi-Zhen Chen, Shin-Hau Chiou, Chia-Ying Chang

Abstract:

This study presents the development of an AI-driven expert system for evaluating the appearance and ripeness of Atemoya fruit. A comprehensive fruit appearance database was constructed by collecting images focused on surface defects and ripeness classification. Deep learning techniques were employed to detect fruit locations in real time, enabling automated assessment of surface imperfections and maturity. Additionally, hyperspectral imaging was utilized to analyze light reflection across different spectral bands, identifying key wavelengths associated with pectin softening in post-ripened fruit. A large-scale multispectral image collection and data analysis process was conducted to establish a robust Atemoya fruit database. The dataset includes high-resolution color images, hyperspectral images covering the 377–1020 nm spectrum, and multispectral images at five specific wavelengths (450, 500, 670, 720, and 800 nm). In total, 4,896 labeled images with ground truth annotations were collected, along with hyperspectral scans of 26 Atemoya fruits (each containing 520 images) and multispectral data from 168 Atemoya fruits (each containing five images). These findings contribute to the advancement of precision agriculture by enhancing fruit quality assessment and post-harvest management.

Keywords: Deep learning, hyperspectral imaging, fruit quality assessment, ripeness detection.

References:

[1] Zhan, Chunyi, et al. "Detection of Apple Sucrose Concentration Based on Fluorescence Hyperspectral Image System and Machine Learning." Foods, vol. 13, no. 22, 2024, p. 3547, https://doi.org/10.3390/foods13223547.
[2] Yang, Xiang, et al. "Non-Destructive Maturity Evaluation of Annona squamosa Using Hyperspectral Imaging and Machine Learning." Journal of Spectral Analysis, vol. 35, no. 4, 2023, pp. 217-228.
[3] Fernandez-Alonso, Francisco Javier, et al. “A Cost-Effective Portable Multiband Spectrophotometer for Precision Agriculture.” Agriculture, vol. 13, no. 8, 2023, pp. 1467-1485, doi:10.3390/agriculture13081467.
[4] Khodabakhshian Kargar, Rasool, et al. “Determining Quality and Maturity of Pomegranates Using Multispectral Imaging.” Journal of the Saudi Society of Agricultural Sciences, vol. 16, no. 4, Elsevier, Oct. 2017, pp. 322–31. DOI: 10.1016/j.jssas.2015.10.004.
[5] Alexey Bochkovskiy, et al. "YOLOv4: Optimal Speed and Accuracy of Object Detection," arXiv 2020, arXiv:2004.10934.