AI-based postharvest management of fruits and vegetables: Enhancing quality, reducing losses, and improving supply chain efficiency

Introduction

Fruits and vegetables are an integral part of a healthy diet, offering a variety of vitamins, minerals, dietary fibre, antioxidants and bioactive compounds. Though nutritional and economical significance, they are one of the most vulnerable commodities of agriculture after harvest. The fresh horticultural produce keeps respiring after harvest, it loses water, softens, changes colour, suffers physiological disorders and is susceptible to microbial decay unlike other grains/dry commodities. These biological and physical phenomena are important determinants of postharvest management and, consequently, market value, consumer acceptance, farmer income and food availability. The post harvest loss is not simply a technical issue of storage. It stands for unused land, water, labour, fertilizer, energy, packaging and transport capacity. It also decreases the supply of nutritious food, and also compromises the livelihood of the producers and small traders. According to Food and Agriculture Organization (FAO) estimates, food losses from post-harvest losses and before reaching the retail market are estimated at 13.3% at the commodity level in 2023, with fruits and vegetables reporting the highest losses at 25.4%. These are indicative figures and provide a clear idea that many horticultural supply chains in some parts of the world would benefit from better management methods, more precise and timely. The traditional methods of post harvest quality assessment rely on visual inspection, sensory assessment, manual grading and laboratory tests.

These strategies are still relevant, however, they have limitations in practice. Manual inspection is subjective and can be biased by fatigue. Lab analysis is accurate but destructive, slow and not suitable for commercial, high throughput sorting. Temperature or shelf-life requirements may not capture the reality of produce in dynamic supply chains. Consequently, losses can take place despite the formal existence of traditional quality-control. Artificial Intelligence provides a new pathway to post harvest transformation. AI systems can analyze images, spectral signals, environmental sensor records, logistics data, and historical quality records to classify the produce, identify defects, assess internal quality, determine the remaining shelf life and make operational decisions. AI can be combined with Internet of Things (IoT) sensors and digital supply-chain platforms, transitioning postharvest management from reaction to action in advance of the inspection. We concentrate here on three aspects of the postharvest management of fruits and vegetables that can be improved by using AI: understanding quality assessment, mitigating postharvest losses, and enhancing supply chain efficiency by predictive and data-driven decision-making.

Literature Review 2.1 Postharvest Losses in Fruits and Vegetables

Post harvest losses in fruits and vegetables are experienced at harvesting, handling in the field, sorting, washing, packing, storage, transport, wholesale markets, retail outlets and consumption level. Losses can be quantitative (weight loss, losses of produce) or qualitative (loss of firmness, colour, flavour, nutritional value and market grade). Fresh produce is metabolically active after harvest, and can deteriorate rapidly if handling conditions are not controlled. The degrees of the temperature is one of the most important factors affecting post harvest losses. The higher the temperature, the more quickly the respiration and ethylene production, loss of water, enzymatic reaction and microbial growth occur. On the other hand, very low temperature can result in chilling injury for sensitive commodities such as banana, mango, cucumber and tomato. The rate of deterioration is also affected by relative humidity, gas composition, packaging and mechanical stresses. However, post harvest management should incorporate biological, physical, environmental and logistic aspects of quality and not focus on quality as a single visual parameter.

Traditional methods of post harvest management

Traditional post harvest management involves careful harvest, maturity indexing, sorting and grading, washing and disinfection, packaging, cold storage, controlled atmosphere storage, modified atmosphere packaging and edible coating and transportation management. When done correctly these practices can minimise losses but require correctly timed practices, skilled staff, infrastructure and ongoing monitoring. A lot of judgments in the fragmented supply chain remain subjective, rather than data-driven. Lab tests can be used to determine firmness, soluble solids content, acidity, moisture content, respiration rate, ethylene production, microbial content and biochemical changes. But in the laboratory, the methods are time consuming, destructive and costly. The requirements of a modern postharvest system are fast, non-destructive and scalable tools that can be used in the farm, packhouse, storage facility, transport vehicle and retail environment.

To explore the role of Artificial Intelligence in

Postharvest Systems: AI is a set of computational methods that can learn from data, find patterns, make predictions, and assist in decision-making. AI has been widely used in postharvest systems, such as machine learning, deep learning, computer vision, sensor fusion, predictive modelling and optimization algorithms. Machine learning models such as random forest, support vector machine, artificial neural networks, gradient boosting and regression models are used to classify quality grades, estimate shelf life, predict firmness, detect spoilage and forecast market demand. Deep learning models, particularly convolutional neural networks, are widely employed in the image-based defect detection, ripeness classification, bruise identification and automated sorting. Time-series data from cold-chain sensors can be analysed by recurrent neural networks and long shortterm memory models to forecast the deterioration of the quality of the product over time. Recent reviews also highlight explainable AI and lightweight models for 'real world' application in post harvest settings with limited resources.

Computer Vision and Image-Based Quality

Assessment: Computer vision is one of the most easily available AI tools for assessing fruit and vegetable quality. The external characteristics of colour, size, shape, surface defects, texture, bruises, disease spots, shrivelling and maturity indicators can be captured by digital cameras. Such features can be used for classification of produce into market grades using deep learning models. Systems based on image can be used in apple defect detection, banana ripeness classification, tomato colour maturity assessment, citrus size grading and wilting detection in leafy vegetables. The main benefits of computer vision is that it is relatively inexpensive, quick and can be automated. It can make the labor dependent lower and make the grading more consistent. But RGB imaging can't see internal defects, early fungal infection, hidden bruising or biochemical changes in quality. Model accuracy can be reduced due to lighting variations, camera angles, background noise, cultivar differences and dust. Computer vision will thus require accompanying imaging conditions, massive data set and complementary sensing technologies.

Hyperspectral Imaging, Spectroscopy and Sensor

Fusion: Hyperspectral imaging is the combination of imaging and spectroscopy, taking information over many wavelengths. It can detect minute differences in chemical and structural changes which are undetectable to the human eye. For fruits and vegetables hyperspectral imaging has been applied to firmness, soluble solids content, acidity, dry matter, maturity, chilling injury, bruising, decay, contamination. Other freshness and spoilage indicators, such as near-infrared spectroscopy, Raman spectroscopy, fluorescence imaging, thermal imaging and electronic nose systems, are also non-destructive. Multiple sensor fusion allows for prediction to be better. It is possible to include temperature, relative humidity, ethylene, carbon dioxide, vibration, position, image features and spectral signals in a shelf-life model. The value of such a multimodal modelling is that the action of different biological and environmental factors interact during postharvest deterioration. In such cases, the use of AI models is beneficial since they can identify non-linear relationships between the variables.

AI to predict the shelf-life of food products.AI predicts the shelf-life of food products

The ability of AI to predict shelf-life is one of the most beneficial applications in post-harvest management. Shelflife estimation can aid producers, wholesalers and retailers in determining when to sell, transport, discount or process and/or consume produce, and can assist the consumer in determining when to eat produce. Shelf-life models are usually classical, where the temperature is assumed as constant and the rate of deterioration is fixed. In actual supply chains, production can also be subject to fluctuations in temperature, relative humidity, delays, mechanical shocks and initial quality variations. Shelf-life models based on AI can continuously update the predictions based on sensor data. The model can also provide a lower remaining shelf-life estimate if temperature abuse is noted during transportation and suggest accelerated distribution. The model could also indicate a longer storage or transit to far market if produce is stored under the optimum condition. This dynamic approach enhances inventory turnover, minimizes waste and increases consumer satisfaction.

AI in Cold Chain and Supply Chain Efficiency

AI can optimize supply-chain processes to better handle routes, predict demand, plan inventories, price based on quality, track shelf-life (dynamic), and manage cold-chain monitoring. Information from IoT sensors can be streamed from storage rooms and transportation vehicles. By detecting temperature variations, AI algorithms can warn managers and identify spoilage risks before they happen. This will allow first-expired-first-out systems that are based on actual remaining shelf life and not fixed arrival dates. AI can also improve supply-chain efficiency by connecting quality data with the demands of the market. Fruits and vegetables that have a shorter shelf life can be channeled to the nearby markets, processing units, institutional buyers or to the discount market. Premium markets can be served with higher grade produce. Quality-sensitive distribution can help to minimise losses and enhance overall value recovery.

Methodology

This is a paper that is conducted in a secondary-data-based narrative review and conceptual analytical approach. The methodology aimed to generate a research paper structure suitable for an empirical manuscript with field/experimental or industry level data. The method is literature synthesis and secondary indicator analysis, conceptual model development and illustration by scenario. First, the recent literature of AI-based postharvest management, fruit and vegetable quality assessment, shelflife prediction, hyperspectral imaging, machine learning and deep learning were reviewed, as well as the monitoring and optimization of supply chains based on the Internet of Things (IoT). The review was limited to studies in the areas of food science, agricultural engineering, biosystems engineering, sensor technology and supply-chain management. Second, the extent of the postharvest problem was illustrated by the use of food-loss indicators on the global level. Data for item 12.3.1 was used as it is based on data from the FAO which is internationally recognised to discuss food losses before retail. The data were presented neither as primary experimental results but as context to demonstrate the need for post harvest AI interventions. Third, AI technologies were grouped into the following categories based on the postharvest function: Quality detection, Grading, shelf-life prediction, Storage control, Transport monitoring, and supply-chain decision support. Fourth, a conceptual framework for post-harvest management based on artificial intelligence (AI) was developed, connecting data collection, AI modelling, decision support, operational intervention and performance evaluation. Finally, illustrative scenario tables and graphs were developed to illustrate how a future empirical study could quantify the benefits of AI. The values given are not indicative of actual outcomes. They should be avoided in favour of data collected from controlled storage trials, packing-house experiments, market surveys, expert scoring or systematic-review extraction of data that has been validated prior to submission to a high-impact journal.

Data Presentation and Analytical Results

This section presents global food-loss indicators, a technology classification matrix, a proposed research model and scenario-based AI adoption results. The purpose is to demonstrate how data tables and graphs can be integrated into a full-length manuscript. The FAO tables use reported values, whereas the AI intervention and priority scores are illustrative and should be replaced with primary data for final publication. Manual grading is subject to inspector fatigue, subjective judgement and variation. AI based computer vision systems can use the same classification rules for each sample, which means that there will be greater uniformity of grade and farmers will be more confident of the traders, retailers and consumers. AI can also help to identify defects at an earlier stage. The image-based models can detect surface defects like bruises, cuts, fungal spots, insect damage and colour abnormalities. Spectroscopy and hyperspectral imaging are able to estimate internal quality attributes like firmness, sugar content, dry matter and early decay. These methods enable quality control to be put in place before the product is consumed, which minimizes rejection, complaints and waste.

Loss Reduction through Predictive Management

AI's key impact is shifting from reactive control to predictive management. Traditional systems have a delayed response to detectable spoilage. AI algorithms can identify patterns that indicate potential quality issues before significant degradation occurs. For instance, if the temperature of a cold storage unit repeatedly moves around, the system can anticipate the potential for spoilage and initiate a corrective action. Shelf-life prediction helps make better inventory decisions. Fixed expiry rules are common in the retail sector and may not be a true representation of the produce. AI-powered remaining-shelf-life prediction can help with dynamic pricing, rotation, and a timely redistribution. This is especially true for fruits and vegetables as soon as deterioration starts, the market value of which decreases rapidly.

Supply-Chain Efficiency

AI enhances the efficiency of the supply chain by incorporating quality data into logistics decisions. In traditional systems, routing and inventory are typically conducted based on quantity and distance and delivery time.

These decisions can be further equipped with quality and remaining shelf-life data by AI-enabled systems. This enables the supply of produce based on the estimated quality life.

The quality produce can be forwarded to a long distance premium market, and the moderate quality produce to local markets. Juices, purees, dried goods, sauces, pickles and animal feed can be made instead of waste from produce that is high risk of spoilage. This is an allocation of resources based on the quality, resulting in higher value recovery along the entire supply chain.

Marketing and Planning for the Region

References

1. Brosnan T, Sun DW. Computer vision for improved quality inspection of food products: A review. Journal of Food Engineering. 2004;61(1):3-16.
2. Cubero S, Aleixos N, Moltó E, Gómez-Sanchis J, Blasco J. The progress made in machine vision applications to automatic inspection and quality assessment of fruits and vegetables. Food and Bioprocess Technology. 2011;4:487-504.
3. Food and Agriculture Organization of the United Nations. State of Food and Agriculture 2019: Taking action on food loss and waste. Rome: FAO; 2019.
4. Food and Agriculture Organization of the United Nations. SDG indicator 12.3.1: Food losses worldwide. Rome: FAO; 2026.
5. Gowen AA, O'Donnell CP, Cullen PJ, Downey G, Frias JM. Hyperspectral imaging: A new process analytical tool for food quality and safety assurance. Trends in Food Science & Technology. 2007;18(12):590-598.
6. Lorente D, Aleixos N, Gómez-Sanchis J, Cubero S, García-Navarrete OL, Blasco J. Recent developments and applications of hyperspectral imaging to the assessment of fruit and vegetable quality. Food and Bioprocess Technology. 2012;5:1121-1142.
7. Mahanti NK, Pandiselvam R, Kothakota A, Ishwarya SP, Chakraborty SK, Kumar M, et al. New NDI techniques for fruit damage detection: Image processing and analysis. Trends in Food Science & Technology. 2022;120:418-438.
8. Pathare PB, Opara UL, Al-Said FAJ. A review of colour measurement and analysis of fresh and processed foods. Food and Bioprocess Technology. 2013;6:36-60.
9. Pu YY, Feng YZ, Sun DW. Recent advances in hyperspectral imaging for quality and safety evaluation of fruits and vegetables: A review. Comprehensive Reviews in Food Science and Food Safety. 2015;14(2):176-188.
10. Rashvand M, Ren Y, Sun DW, Senge J, Krupitzer C, Fadiji T, et al. Recent advances and current challenges on using artificial intelligence for shelf-life prediction of different food products. Trends in Food Science & Technology. 2025;159:104989.
11. Sun DW, editor. Food Quality Evaluation Using Computer Vision Technology. 2nd ed. London: Academic Press; 2016.
12. Vignati S, Tugnolo A, Giovenzana V, Pampuri A, Casson A, Guidetti R, et al. Basic concepts and applications for quality assessment of fresh-cut fruit and vegetable products by hyperspectral imaging. Applied Sciences. 2023;13(17):9740.
13. Wieme J, Mollazade K, Malounas I, Zude-Sasse M, Zhao M, Gowen A, et al. Application of hyperspectral imaging systems and artificial intelligence for quality assessment of fruit, vegetables and mushrooms: A review. Biosystems Engineering. 2022;222:156-176.
14. Zhang B, Huang W, Li J, Zhao C, Fan S, Wu J, et al. Review of principles, developments and applications of computer vision in external quality inspection of fruits and vegetables. Food Research International. 2014;62:326-343.
15. Zude-Sasse M, Fountas S, Gemtos TA, Abu-Khalaf N. Use of precision agriculture in horticultural crops. European Journal of Horticultural Science. 2016;81(2):78-90.