Applications of machine learning techniques for enhancing nondestructive food quality and safety detection

Interpretation and explanation of computer vision classification of carambola (Averrhoa carambola L.) according to maturity stage

The classification of carambola, also known as starfruit, according to quality parameters is usually conducted by trained human evaluators through visual inspections. This is a costly and subjective method that can generate high variability in results. As an alternative, computer vision systems (CVS) combined with deep learning (DCVS) techniques have been introduced in the industry as a powerful and an innovative tool for the rapid and non-invasive classification of fruits. However, validating the learning capability and trustworthiness of a DL model, aka black box, to obtain insights can be challenging. To reduce this gap, we propose an integrated eXplainable Artificial Intelligence (XAI) method for the classification of carambolas at different maturity stages. We compared two Residual Neural Networks (ResNet) and Visual Transformers (ViT) to identify the image regions that are enhanced by a Random Forest (RF) model, with the aim of providing more detailed information at the feature level for classifying the maturity stage. Changes in fruit colour and physicochemical data throughout the maturity stages were analysed, and the influence of these parameters on the maturity stages was evaluated using the Gradient-weighted Class Activation Mapping (Grad-CAM), the Attention Maps using RF importance. The proposed approach provides a visualization and description of the most important regions that led to the model decision, in wide visualization follows the models an importance features from RF. Our approach has promising potential for standardized and rapid carambolas classification, achieving 91 % accuracy with ResNet and 95 % with ViT, with potential application for other fruits.

Spectral intelligent detection for aflatoxin B1 via contrastive learning based on Siamese network

Aflatoxins, harmful substances found in peanuts, corn, and their derivatives, pose significant health risks. Addressing this, the presented research introduces an innovative MSGhostDNN model, merging contrastive learning with multi-scale convolutional networks for precise aflatoxin detection. The method significantly enhances feature discrimination, achieving an impressive 97.87% detection accuracy with a pre-trained model. By applying Grad-CAM, it further refines the model to identify key wavelengths, particularly 416 nm, and focuses on 40 key wavelengths for optimal performance with 97.46% accuracy. The study also incorporates a task dimensionality reduction approach for continuous learning, allowing effective ongoing aflatoxin spectrum monitoring in peanuts and corn. This approach not only boosts aflatoxin detection efficiency but also sets a precedent for rapid online detection of similar toxins, offering a promising solution to mitigate the health risks associated with aflatoxin exposure.

Elevating nanomaterial optical sensor arrays through the integration of advanced machine learning techniques for enhancing visual inspection of food quality and safety

Food quality and safety problems caused by inefficient control in the food chain have significant implications for human health, social stability, and economic progress and optical sensor arrays (OSAs) can effectively address these challenges. This review aims to summarize the recent applications of nanomaterials-based OSA for food quality and safety visual monitoring, including colourimetric sensor array (CSA) and fluorescent sensor array (FSA). First, the fundamental properties of various advanced nanomaterials, mainly including metal nanoparticles (MNPs) and nanoclusters (MNCs), quantum dots (QDs), upconversion nanoparticles (UCNPs), and others, were described. Besides, the diverse machine learning (ML) and deep learning (DL) methods of high-dimensional data obtained from the responses between different sensing elements and analytes were presented. Moreover, the recent and representative applications in pesticide residues, heavy metal ions, bacterial contamination, antioxidants, flavor matters, and food freshness detection were comprehensively summarized. Finally, the challenges and future perspectives for nanomaterials-based OSAs are discussed. It is believed that with the advancements in artificial intelligence (AI) techniques and integrated technology, nanomaterials-based OSAs are expected to be an intelligent, effective, and rapid tool for food quality assessment and safety control.

Advanced detection tools in food fraud: A systematic review for holistic and rational detection method based on research and patents

Modern food chain supply management necessitates the dire need for mitigating food fraud and adulterations. This holistic review addresses different advanced detection technologies coupled with chemometrics to identify various types of adulterated foods. The data on research, patent and systematic review analyses (2018-2023) revealed both destructive and non-destructive methods to demarcate a rational approach for food fraud detection in various countries. These intricate hygiene standards and AI-based technology are also summarized for further prospective research. Chemometrics or AI-based techniques for extensive food fraud detection are demanded. A systematic assessment reveals that various methods to detect food fraud involving multiple substances need to be simple, expeditious, precise, cost-effective, eco-friendly and non-intrusive. The scrutiny resulted in 39 relevant experimental data sets answering key questions. However, additional research is necessitated for an affirmative conclusion in food fraud detection system with modern AI and machine learning approaches.

Integrating Vision-Language Models for Accelerated High-Throughput Nutrition Screening

Addressing the critical need for swift and precise nutritional profiling in healthcare and in food industry, this study pioneers the integration of vision-language models (VLMs) with chemical analysis techniques. A cutting-edge VLM is unveiled, utilizing the expansive UMDFood-90k database, to significantly improve the speed and accuracy of nutrient estimation processes. Demonstrating a macro-AUCROC of 0.921 for lipid quantification, the model exhibits less than 10% variance compared to traditional chemical analyses for over 82% of the analyzed food items. This innovative approach not only accelerates nutritional screening by 36.9% when tested amongst students but also sets a new benchmark in the precision of nutritional data compilation. This research marks a substantial leap forward in food science, employing a blend of advanced computational models and chemical validation to offer a rapid, high-throughput solution for nutritional analysis.

Application of machine learning approach on halal meat authentication principle, challenges, and prospects: A review

Meat is a source of essential amino acids that are necessary for human growth and development, meat can come from dead, alive, Halal, or non-Halal animal species which are intentionally or economically (adulteration) sold to consumers. Sharia has prohibited the consumption of pork by Muslims. Because of the activities of adulterators in recent times, consumers are aware of what they eat. In the past, several methods were employed for the authentication of Halal meat, but numerous drawbacks are attached to this method such as lack of flexibility, limited application, time,consumption and low level of accuracy and sensitivity. Machine Learning (ML) is the concept of learning through the development and application of algorithms from given data and making predictions or decisions without being explicitly programmed. The techniques compared with traditional methods in Halal meat authentication are fast, flexible, scaled, automated, less expensive, high accuracy and sensitivity. Some of the ML approaches used in Halal meat authentication have proven a high percentage of accuracy in meat authenticity while other approaches show no evidence of Halal meat authentication for now. The paper critically highlighted some of the principles, challenges, successes, and prospects of ML approaches in the authentication of Halal meat.

Unleashing the Potential of Digitalization in the Agri-Food Chain for Integrated Food Systems

Digitalization transforms many industries, especially manufacturing, with new concepts such as Industry 4.0 and the Industrial Internet of Things. However, information technology also has the potential to integrate and connect the various steps in the supply chain. For the food industry, the situation is ambivalent: It has a high level of automatization, but the potential of digitalization is so far not used today. In this review, we discuss current trends in information technology that have the potential to transform the food industry into an integrated food system. We show how this digital transformation can integrate various activities within the agri-food chain and support the idea of integrated food systems. Based on a future-use case, we derive the potential of digitalization to tackle future challenges in the food industry and present a research agenda.

Visible detection of chilled beef freshness using a paper-based colourimetric sensor array combining with deep learning algorithms

This study developed an innovative approach that combines a colourimetric sensor array (CSA) composed of twelve pH-response dyes with advanced algorithms, aiming to detect amine gases and assess the freshness of chilled beef. With the assistance of multivariate statistical analysis, the sensor array can effectively distinguish five amine gases and enable rapid quantification of trimethylamine vapour with a limit of detection (LOD) of 8.02 ppb and visually monitor the fresh levels of chilled beef. Moreover, the utilization of deep learning models (ResNet34, VGG16, and GoogleNet) for chilled beef freshness evaluation achieved an overall accuracy of 98.0 %. Furthermore, t-distributed stochastic neighbour embedding (t-SNE) visualized the feature extraction process and provided explanations to understand the classification process of deep learning. The results demonstrated that applying deep learning techniques in the process of pattern recognition of CSA can help in realizing the rapid, robust, and accurate assessment of chilled beef freshness.

Food authentication, current issues, analytical techniques, and future challenges: A comprehensive review

Food authentication and contamination are significant concerns, especially for consumers with unique nutritional, cultural, lifestyle, and religious needs. Food authenticity involves identifying food contamination for many purposes, such as adherence to religious beliefs, safeguarding health, and consuming sanitary and organic food products. This review article examines the issues related to food authentication and food fraud in recent periods. Furthermore, the development and innovations in analytical techniques employed to authenticate various food products are comprehensively focused. Food products derived from animals are susceptible to deceptive practices, which can undermine customer confidence and pose potential health hazards due to the transmission of diseases from animals to humans. Therefore, it is necessary to employ suitable and robust analytical techniques for complex and high-risk animal-derived goods, in which molecular biomarker-based (genomics, proteomics, and metabolomics) techniques are covered. Various analytical methods have been employed to ascertain the geographical provenance of food items that exhibit rapid response times, low cost, nondestructiveness, and condensability.

Seabream quality monitoring throughout the supply chain using a portable multispectral imaging device

Monitoring food quality throughout the supply chain in a rapid and cost-effective way allows on-time decision making, reducing food waste and increasing sustainability. In that framework, a portable multispectral imaging sensor was used, while the acquired data in combination with neural networks were evaluated for the prediction of fish fillets quality. Images of fish fillets were acquired using samples from both aquaculture and retail stores of different packaging and fish parts. The obtained products (air or vacuum packaged) were further stored at different temperature conditions. In parallel to image acquisition, microbial quality was estimated as well. The data were used for the training of predictive neural models that aimed to estimate total aerobic counts (TAC). The models were developed and validated using data from aquaculture and were externally validated with samples purchased from the retail stores. The set up allowed the evaluation of models for the different parts of the fish and conditions. The performance for the validation set was similar for flesh (RMSE: 0.402-0.547) and skin side (RMSE: 0.500-0.533) of the fish fillets. The performance for the different packaging conditions was also similar, however, in the external validation, the vacuum-packaged samples showed better performance in terms of RMSE compared to the air-packaged ones. Models irrespective of packaging condition are very important for cases where the products' history is unknown although the prediction capability was not as high as in the models per packaging condition individually. The models tested with unknown samples (i.e., from retail stores) showed poorer performance (RMSE: 1.061-1.414) compared to the models validated with data partitioning (RMSE: 0.402-0.547). Multispectral imaging sensor appeared to be efficient for the rapid assessment of the microbiological quality of fish fillets for all the different cases evaluated.

Machine-learning-driven optical immunosensor based on microspheres-encoded signal transduction for the rapid and multiplexed detection of antibiotics in milk

Antibiotic residues are the most common contaminants in milk and other related dairy products. Simultaneous, convenient, and stable detection of antibiotic residues in foods is vital to secure public health. Herein, we proposed an optical immunosensor with easily-functionalized polystyrene nanoparticles differing in size and quantity, and bearing multiplex signal probes for the simultaneous detection of multiple antibiotics through a simple one-step signal conversion reaction. After the integration of the machine-learning-based transcoding analysis, this sensor is suitable for multiplexed detection of antibiotics in a broad linear range from pg/mL to ng/mL within 30 min, with an overall accuracy of >99 %. Compared to the conventional standard chemiluminescence immunoassays, this immunosensor is suitable for the accurate quantification of multiple antibiotics in milk, with improved accuracy, reduced costs, and simplified procedure. This ensures its applications in food safety monitoring when simultaneous detection of multiple hazardous substances in food matrices is needed.

Combined quantitative lipidomics and back-propagation neural network approach to discriminate the breed and part source of lamb

The study successfully utilized an analytical approach that combined quantitative lipidomics with back-propagation neural networks to identify breed and part source of lamb using small-scale samples. 1230 molecules across 29 lipid classes were identified in longissimus dorsi and knuckle meat of both Tan sheep and Bahan crossbreed sheep. Applying multivariate statistical methods, 12 and 7 lipid molecules were identified as potential markers for breed and part identification, respectively. Stepwise linear discriminant analysis was applied to select 3 and 4 lipid molecules, respectively, for discriminating lamb breed and part sources, achieving correct rates of discrimination of 100 % and 95 %. Additionally, back-propagation neural network proved to be a superior method for identifying sources of lamb meat compared to other machine learning approaches. These findings indicate that integrating lipidomics with back-propagation neural network approach can provide an effective strategy to trace and certify lamb products, ensuring their quality and protecting consumer rights.

Antimicrobial peptides: An alternative to traditional antibiotics

As antibiotic-resistant bacteria and genes continue to emerge, the identification of effective alternatives to traditional antibiotics has become a pressing issue. Antimicrobial peptides are favored for their safety, low residue, and low resistance properties, and their unique antimicrobial mechanisms show significant potential in combating antibiotic resistance. However, the high production cost and weak activity of antimicrobial peptides limit their application. Moreover, traditional laboratory methods for identifying and designing new antimicrobial peptides are time-consuming and labor-intensive, hindering their development. Currently, novel technologies, such as artificial intelligence (AI) are being employed to develop and design new antimicrobial peptide resources, offering new opportunities for the advancement of antimicrobial peptides. This article summarizes the basic characteristics and antimicrobial mechanisms of antimicrobial peptides, as well as their advantages and limitations, and explores the application of AI in antimicrobial peptides prediction amd design. This highlights the crucial role of AI in enhancing the efficiency of antimicrobial peptide research and provides a reference for antimicrobial drug development.

Sniffing Like a Wine Taster: Multiple Overlapping Sniffs (MOSS) Strategy Enhances Electronic Nose Odor Recognition Capability

As highly promising devices for odor recognition, current electronic noses are still not comparable to human olfaction due to the significant disparity in the number of gas sensors versus human olfactory receptors. Inspired by the sniffing skills of wine tasters to achieve better odor perception, a multiple overlapping sniffs (MOSS) strategy is proposed in this study. The MOSS strategy involves rapid and continuous inhalation of odorants to stimulate the sensor array to generate feature-rich temporal signals. Computational fluid dynamics simulations are performed to reveal the mechanism of complex dynamic flows affecting transient responses. The proposed strategy shows over 95% accuracy in the recognition experiments of three gaseous alkanes and six liquors. Results demonstrate that the MOSS strategy can accurately and easily recognize odors with a limited sensor number. The proposed strategy has potential applications in various odor recognition scenarios, such as medical diagnosis, food quality assessment, and environmental surveillance.

Prediction of the Tribological Properties of Polytetrafluoroethylene Composites Based on Experiments and Machine Learning

Because of the complex nonlinear relationship between working conditions, the prediction of tribological properties has become a difficult problem in the field of tribology. In this study, we employed three distinct machine learning (ML) models, namely random forest regression (RFR), gradient boosting regression (GBR), and extreme gradient boosting (XGBoost), to predict the tribological properties of polytetrafluoroethylene (PTFE) composites under high-speed and high-temperature conditions. Firstly, PTFE composites were successfully prepared, and tribological properties under different temperature, speed, and load conditions were studied in order to explore wear mechanisms. Then, the investigation focused on establishing correlations between the friction and wear of PTFE composites by testing these parameters through the prediction of the friction coefficient and wear rate. Importantly, the correlation results illustrated that the friction coefficient and wear rate gradually decreased with the increase in speed, which was also proven by the correlation coefficient. In addition, the GBR model could effectively predict the tribological properties of the PTFE composites. Furthermore, an analysis of relative importance revealed that both load and speed exerted a greater influence on the prediction of the friction coefficient and wear rate.

Spectral data fusion in nondestructive detection of food products: Strategies, recent applications, and future perspectives

In recent years, the food industry has shown a growing interest in the development of rapid and nondestructive analytical methods. However, the utilization of a solitary nondestructive detection technique offers only a constrained extent of physical or chemical insights regarding the sample under examination. To overcome this limitation, the amalgamation of spectroscopy with data fusion strategies has emerged as a promising approach. This comprehensive review delves into the fundamental principles and merits of low-level, mid-level, and high-level data fusion strategies within the domain of food analysis. Various data fusion techniques encompassing spectra-to-spectra, spectra-to-machine vision, spectra-to-electronic nose, and spectra-to-nuclear magnetic resonance are summarized. Moreover, this review also provides an overview of the latest applications of spectral data fusion techniques (SDFTs) for classification, adulteration, quality evaluation, and contaminant detection within the purview of food safety analysis. It also addresses current challenges and future prospects associated with SDFTs in real-world applications. Despite the extant technical intricacy, the ongoing evolution of online data fusion platforms and the emergence of smartphone-based multi-sensor fusion detection technology augur well for the pragmatic realization of SDFTs, endowing them with formidable capabilities for both qualitative and quantitative analysis in the realm of food analysis.

Detection of Red Pepper Powder Adulteration with Allura Red and Red Pepper Seeds Using Hyperspectral Imaging

This study used shortwave infrared (SWIR) technology to determine whether red pepper powder was artificially adulterated with Allura Red and red pepper seeds. First, the ratio of red pepper pericarp to seed was adjusted to 100:0 (P100), 75:25 (P75), 50:50 (P50), 25:75 (P25), or 0:100 (P0), and Allura Red was added to the red pepper pericarp/seed mixture at 0.05% (A), 0.1% (B), and 0.15% (C). The results of principal component analysis (PCA) using the L, a, and b values; hue angle; and chroma showed that the pure pericarp powder (P100) was not easily distinguished from some adulterated samples (P50A-C, P75A-C, and P100B,C). Adulterated red pepper powder was detected by applying machine learning techniques, including linear discriminant analysis (LDA), linear support vector machine (LSVM), and k-nearest neighbor (KNN), based on spectra obtained from SWIR (1,000-1,700 nm). Linear discriminant analysis determined adulteration with 100% accuracy when the samples were divided into four categories (acceptable, adulterated by Allura Red, adulterated by seeds, and adulterated by seeds and Allura Red). The application of SWIR technology and machine learning detects adulteration with Allura Red and seeds in red pepper powder.

Recent advances on natural colorants-based intelligent colorimetric food freshness indicators: fabrication, multifunctional applications and optimization strategies

With the increasing concerns of food safety and public health, tremendous efforts have been concentrated on the development of effective, reliable, nondestructive methods to evaluate the freshness level of different kinds of food. Natural colorants-based intelligent colorimetric indicators which are typically constructed with natural colorants and polymer matrices has been regarded as an innovative approach to notify the customers and retailers of the food quality during the storage and transportation procedure in real-time. This review briefly elucidates the mechanism of natural colorants used for intelligent colorimetric indicators and fabrication methodologies of natural colorants-based food freshness indicators. Subsequently, their multifunctional applications in intelligent food packaging systems like antioxidant packaging, antimicrobial packaging, biodegradable packaging, UV-blocking packaging and inkless packaging are well introduced. This paper also summarizes several optimizing strategies for the practical application of this advanced technology from different perspectives. Strategies like adopting a hydrophobic matrix, constructing double-layer film and encapsulation have been developed to improve the stability of the indicators. Co-pigmentation, metal ion complexation, pigment-mixing and using substrates with high surface area are proved to be effective to enhance the sensitivity of the indicators. Approaches include multi-index evaluation, machine learning and smartphone-assisted evaluation have been proven to improve the accuracy of the intelligent food freshness indicators. Finally, future research opportunities and challenges are proposed. Based on the fundamental understanding of natural colorants-based intelligent colorimetric food freshness indicators, and the latest research and findings from literature, this review article will help to develop better, lower cost and more reliable food freshness evaluation technique for modern food industry.

A Comprehensive Review on the Excitation-Emission Matrix Fluorescence Spectroscopic Characterization of Petroleum-Containing Substances: Principles, Methods, and Applications

Petroleum-containing substance (PCS) is a general term used for petroleum and its derivatives. A comprehensive characterization of PCSs is crucial for resource exploitation, economic development and environmental protection. Fluorescence spectroscopy, especially excitation-emission matrix fluorescence (EEMF) spectroscopy, has been proved to be a powerful tool to characterize PCSs since its remarkable sensitivity, selectivity, simplicity and high efficiency. However, there is a lack of systematic review focusing on this field in the literature. This paper reviews the fundamental principles and measurements of EEMF for characterizing PCSs, and makes a systematic introduction to various information mining methods including basic peak information extraction, spectral parameterization and some commonly used chemometric methods. In addition, recent advances in applying EEMF to characterize PCSs during the whole life-cycle process of petroleum are also revisited. Furthermore, the current limitations of EEMF in the measurement and characterization of PCSs are discussed and corresponding solutions are provided. For promoting the future development of this field, the urgent need to build a relatively complete EEMF fingerprint library to trace PCSs, not only pollutants but also crude oil and petroleum products, is proposed. Finally, the extensions of EEMF to high-dimensional chemometrics and deep learning are prospected, with the expectation of solving more complex systems and problems.

Multiscale Deepspectra Network: Detection of Pyrethroid Pesticide Residues on the Hami Melon

The problem of pyrethroid residues has become a topical issue, posing a potential food safety concern. Pyrethroid pesticides are widely used to prevent and combat pests in Hami melon cultivation. Due to its high sensitivity and accuracy, gas chromatography (GC) is used most frequently for detecting pyrethroid pesticide residues. However, GC has a high cost and complex operation. This study proposed a deep-learning approach based on the one-dimensional convolutional neural network (1D-CNN), named Deepspectra network, to detect pesticide residues on the Hami melon based on visible/near-infrared (380-1140 nm) spectroscopy. Three combinations of convolution kernels were compared in the single-scale Deepspectra network. The convolution group of "5 × 1" and "3 × 1" kernels obtained a better overall performance. The multiscale Deepspectra network was compared to three single-scale Deepspectra networks on the preprocessing spectral data and obtained better results. The coefficient of determination (R2) for lambda-cyhalothrin and beta-cypermethrin was 0.758 and 0.835, respectively. The residual predictive deviation (RPD) for lambda-cyhalothrin and beta-cypermethrin was 2.033 and 2.460, respectively. The Deepspectra networks were compared with two conventional regression models: partial least square regression (PLSR) and support vector regression (SVR). The results showed that the multiscale Deepspectra network outperformed the other models. It was found that the multiscale Deepspectra network could be a novel approach for the quantitative estimation of pyrethroid pesticide residues on the Hami melon. These findings can also provide an effective strategy for spectral analysis.

Emerging technologies for preservation and quality evaluation of postharvest edible mushrooms: A review

Edible mushrooms are the highly demanded foods of which production and consumption have been steadily increasing globally. Owing to the quality loss and short shelf-life in harvested mushrooms, it is necessary for the implementation of effective preservation and intelligent evaluation technologies to alleviate this issue. The aim of this review was to analyze the development and innovation thematic lines, topics, and trends by bibliometric analysis and review of the literature methods. The challenges faced in researching these topics were proposed and the mechanisms of quality loss in mushrooms during storage were updated. This review summarized the effects of chemical processing (antioxidants, ozone, and coatings), physical treatments (non-thermal plasma, packaging and latent thermal storage) and other emerging application on the quality of fresh mushrooms while discussing the efficiency in extending the shelf-life. It also discussed the emerging evaluation techniques based on the various chemometric methods and computer vision system in monitoring the freshness and predicting the shelf-life of mushrooms which have been developed. Preservation technology optimization and dynamic quality evaluation are vital for achieving mushroom quality control. This review can provide a comprehensive research reference for reducing mushroom quality loss and extending shelf-life, along with optimizing efficiency of storage and transportation operations.

Artificial Intelligence in Food Safety: A Decade Review and Bibliometric Analysis

Artificial Intelligence (AI) technologies have been powerful solutions used to improve food yield, quality, and nutrition, increase safety and traceability while decreasing resource consumption, and eliminate food waste. Compared with several qualitative reviews on AI in food safety, we conducted an in-depth quantitative and systematic review based on the Core Collection database of WoS (Web of Science). To discover the historical trajectory and identify future trends, we analysed the literature concerning AI technologies in food safety from 2012 to 2022 by CiteSpace. In this review, we used bibliometric methods to describe the development of AI in food safety, including performance analysis, science mapping, and network analysis by CiteSpace. Among the 1855 selected articles, China and the United States contributed the most literature, and the Chinese Academy of Sciences released the largest number of relevant articles. Among all the journals in this field, PLoS ONE and Computers and Electronics in Agriculture ranked first and second in terms of annual publications and co-citation frequency. The present character, hot spots, and future research trends of AI technologies in food safety research were determined. Furthermore, based on our analyses, we provide researchers, practitioners, and policymakers with the big picture of research on AI in food safety across the whole process, from precision agriculture to precision nutrition, through 28 enlightening articles.

Analyzing Volatile Compounds of Young and Mature Docynia delavayi Fruit by HS-SPME-GC-MS and rOAV

This study focused on the examination of the volatile compounds and fragrance properties of the young and mature fruit of Docynia delavayi. Headspace solid-phase microextraction combined with gas chromatography−mass spectrometry (HS-SPME-GC-MS) was applied for identifying 42 volatile compounds, with young and mature fruit containing 36 and 42 compounds, respectively. Heat map cluster analysis, principal component analysis (PCA), orthogonal partial least squares discriminant analysis (OPLS-DA), and independent sample t-testing were used to analyze sample differences. Based on a variable importance in projection (VIP) > 1 and p < 0.05, 23 key volatile compounds such as octanal, geranylacetone, butyl acetate, and dihydro-β-ionone were screened. β-Ionone and phenethyl acetate made the largest contribution to the aroma of D. delavayi after analyzing the relative odor activity value (rOAV) of the key volatile compounds and their aroma descriptors. Young D. delavayi fruit exhibited a prominent woody scent, while mature D. delavayi fruit had more intense floral and rosy aromas. The findings may lay a foundation for comprehensively developing and utilizing D. delavayi fruit.