Abstract: Non-destructive testing of fruit ripeness is of great significance for reducing losses during fruit production, transportation, and storage. Among traditional methods, empirical judgment is highly subjective and less accurate, while destructive testing compromises the integrity of samples. Both methods suffer from low detection efficiency. In recent years, with the rapid development of non-destructive testing technologies such as near-infrared spectroscopy, hyperspectral imaging, machine vision, acoustic vibration, electrical properties, tactile sensing, and electronic noses, the key parameters highly correlated with fruit ripeness-such as color, hardness, sugar-acid ratio, and volatile gases-have been used to accurately assess the maturity state of fruits through machine learning approaches. This study begins with the physiological changes during fruit ripening and systematically reviews the fundamental principles and applications of non-destructive testing technologies. Furthermore, it summarizes current challenges in the field, including environmental interference, lack of data standards, and the balance between cost, complexity, and multimodal integration. The study also explores future directions, such as the development of low-cost multimodal sensor technologies, the establishment of data standards, and improving the interpretability of machine learning.

Key words: non-destructive testing, fruit ripeness, multimodal fusion, sensors, machine learning