A systematic review of the internet of things and artificial intelligence applications in milk quality monitoring and analysis

Background: Milk quality analysis (MQA) is critical for ensuring high-quality dairy production and enhancing the productivity of dairy processing plants. The integration of Internet of Things (IoT) and artificial intelligence (AI) can revolutionise MQA by enabling automated, real-time detection and monitoring of key parameters (i.e. composition, adulterants, indicators of microbial quality), reducing human error and improving decision-making in the dairy supply chain. Aim(s): This systematic review aimed to evaluate the application of IoT and AI in MQA over the past decade, focussing on three critical areas: composition monitoring, adulteration detection and indicators of microbial quality. Methods: A systematic analysis of 43 studies was conducted, comparing sensing technologies (e.g. spectroscopy and gas sensors), data communication protocols (i.e. WifI, Bluetooth), data processing strategies (edge computing, cloud platforms) and AI models (Machine Learning, Deep Learning). The methodologies implemented, along with model performance evaluation metrics and the scalability of the proposed technologies, will be assessed. Major Findings: Spectroscopy emerged as a leading method for compositional and adulteration analysis (62%), while gas sensors were preferred for microbial spoilage detection. Edge computing was the leading data processing strategy (64%), enabling real-time analysis, though hybrid edge–cloud platforms demonstrated good scalability potential. Machine Learning models (e.g. Support Vector Machines, Random Forest) achieved high predictive accuracy, while Deep Learning excelled in classification tasks, particularly for adulteration detection. Scientific Implications: Despite the significant advancements, challenges remain in developing generalised models applicable across diverse milk samples (geographic regions and milk species). Current systems face limitations in data communication from remote farms with poor connectivity and computational constraints when processing complex spectral data on edge devices. Future advancements in 5G networks could address latency issues in large-scale deployments, while smart sensors with embedded preprocessing capabilities may reduce computational constraints. Hybrid edge-cloud platforms enable real-time analysis while supporting model updates, enhancing smart dairy monitoring.