Insights into recent developments and obstacles in automated fruit ripeness classification

The ripening stage of a fruit is a crucial determinant of its quality. Traditional assessment methods for fruit ripeness are often subjective, error-prone, and time-consuming. Therefore, numerous attempts are being made to establish rapid and reliable techniques for assessing the ripeness of fruits and vegetables. Innovative technologies, such as electronic noses, imaging technologies, and nondestructive firmness sensors, have been employed to evaluate fruit maturity based on various descriptive criteria. The most effective approaches are dominated by artificial intelligence (AI) strategies, namely, machine learning (ML), deep learning (DL), and most recently explainable artificial intelligence (XAI). With the rapid evolution of automated ripeness classification, this review summarizes recent developments in sensors, algorithms, and real-world applications across fruits and vegetables. It also evaluates the main obstacles (environmental variability, generalizability, and data quality) and outlines promising avenues for feasible, scalable solutions. The findings of this study emphasize the need for standardized methodologies to enable reproducibility and outcome comparisons. Developing a common framework for evaluating various algorithms, sensing systems, and comparable datasets as well as standardized evaluation criteria could significantly advance automated fruit ripeness classification. Higher classification accuracy can be achieved using more complex deep learning algorithms, such as convolutional neural networks, despite their high accuracy, the lack of transparency and explainability limits the broader industrial adoption of such techniques. Explainable artificial intelligence (XAI) is emerging as a promising approach, enhancing transparency and trust by addressing the interpretability challenges inherent in complex AI-driven models. In conclusion, although automated ripeness classification has made significant advances in recent years, improvements in data processing, standardization, and accuracy are still needed to address current challenges. By integrating appropriate sensing technologies and selecting suitable feature data and algorithms, automatic classification of fruit ripeness can be continually enhanced, thereby improving classification accuracy. This progress will ultimately contribute to advancements in agricultural intelligence, supply chain management, higher food quality, and reduced waste generation. • Provides a comprehensive review of deep learning applications in fruit ripening detection. • Explores recent advances in sensor technologies for non-destructive fruit maturity monitoring. • Summarizes the effectiveness of techniques such as imaging and electronic noses in evaluating fruit properties. • Discusses the strengths and limitations of current approaches for fruit ripeness classification. • Identifies future research directions and potential challenges in automated fruit quality assessment.