A comprehensive review of quality measurements of fruits using electronic nose and computer vision

The quality of fruits is crucial for consumer demand, market acceptance and environmental health. Many studies have measured and predicted fruit quality using electronic nose (E-nose) and computer vision. However, reviews simultaneously covering both E-nose and computer vision research on fruit quality is scarce. A review of both E-nose and computer vision technologies is critical for advancing research and guiding practical applications in fruit quality assessment. Its primary objective is to provide an in-depth discussion of findings, limitations, and knowledge gaps, while highlighting the relevance of this field to environmental health. The review found that apples and strawberries were the most frequently studied fruits. Metal-oxide semiconductor (MOS) sensors were the most widely used in E-nose applications, with principal component analysis (PCA), linear discriminant analysis (LDA), and support vector machines (SVM) as the primary classification methods. In computer vision studies, RGB and hyperspectral imaging (HSI) were widely used, and White and Dark Reference Image Correction was the most common preprocessing method. PCA was the most frequently used machine learning method, followed by partial least squares discriminant analysis (PLS-DA), SVM, and convolutional neural networks (CNN). Research mainly focused on a few fruits, leaving many widely consumed types understudied. Existing methods were limited by sensor performance, small datasets, weak validation, little real-world testing, and little consideration of shelf life. We conclude that integrating suitable preprocessing with combined E-nose and computer vision techniques improves fruit quality classification, while accounting for shelf life and exploring environmental implications remain essential directions for future research.