Abstract: Fruit quality is one of the most important indicators in the modern fruit industry. Accurate quality evaluation is often required for national product safety in modern agriculture, enabling high-quality products and overall industrial efficiency. However, conventional evaluation of fruit quality cannot fully meet the rapid, large-scale inspection due to its destructive, labor-intensive, and time-consuming demands. Fortunately, non-destructive detection technologies can be expected to serve as important alternatives in recent decades, because of their rapid response, non-invasive operation, high throughput, and online integration. This review aims to summarize recent advances in spectral- and imaging-based non-destructive detection technologies and their equipment for fruit quality. Eight representative technologies were selected, including visible/near-infrared (Vis/NIR) spectroscopy, hyperspectral imaging (HSI), X-ray imaging, laser speckle imaging, Raman spectroscopy, fluorescence spectroscopy, terahertz (THz) spectroscopy, and machine vision (MV). Sensing, application, and technical trends were then analyzed for each technology. Attention was given to their applications in internal quality assessment, external defect detection, physiological status monitoring, disease and pest identification, safety inspection, and intelligent grading. A comparison was further made on their sensing, penetration depth, spatial resolution, engineering feasibility, suitable application scenarios, and online implementation potential. Different technologies of non-destructive detection exhibited significant variations in information perception dimensions and target characteristics. Therefore, these techniques were complementary rather than mutually substitutive. From the perspective of intelligent fruit quality evaluation, future technologies increasingly rely on multimodal sensing and collaborative information fusion for more accurate and robust quality. Nevertheless, most existing studies were still conducted under controlled laboratory conditions, leading to restricted large-scale industrial application. One major challenge was also the acquisition of high-quality signals. Another limitation was concerned with the construction of robust predictive models. Furthermore, "black box" systems retained less mechanistic interpretability. In addition, efficient fusion of heterogeneous data from different sensing modalities remained insufficient because of different spatial scales, feature representation, and data structure, although multimodal sensing attracted increasing attention. Future research should focus on promising potential directions. 1) Greater emphasis should be placed on the fruit quality formation and the interaction between biological tissues and detection signals, thereby providing a stronger theoretical foundation for non-destructive sensing. 2) The key sensing technologies should be further improved after signal acquisition strategies, high-sensitivity detectors, optimal optical configurations, and advanced interference suppression. 3) Artificial intelligence and data-driven approaches should be integrated into non-destructive detection. The interpretable, lightweight, transferable, and adaptive intelligent models were often required to improve detection accuracy and robustness using deep learning, transfer learning, and multimodal data fusion. 4) Future equipment was used to promote online, portable, miniaturized, and intelligent detection for the stable operation in complex industrial environments. Finally, the standardized databases, unified evaluation criteria, and collaborative industrial frameworks can also be used to accelerate technology transfer and large-scale commercialization. Overall, advanced sensing technologies, intelligent algorithms, and industrial equipment can be integrated to drive the next generation of non-destructive detection of fruit quality. The finding can provide important technical support for intelligent post-harvesting in modern smart agriculture.