Edge-cloud-blockchain integrated framework for agricultural product traceability: AI-driven anomaly detection and smart contract optimization

Introduction With increasing global demands for food safety, transparency, and sustainability, agricultural product traceability systems must achieve both real-time performance and trusted data sharing. However, traditional centralized systems and pure blockchain-based systems still face limitations such as data heterogeneity, cross-domain barriers, high overhead, and low response efficiency. Methods This study proposes an edge-cloud-blockchain integrated framework for agricultural product traceability. The framework combines edge-side real-time preprocessing and anomaly detection, cloud-side global model training through federated learning, Hyperledger Fabric with on-chain/off-chain data partitioning, and AI-enhanced smart contracts for automatic anomaly response. Lightweight TCN-SE and PDCN-ECA models are deployed to support missing-value imputation and drift detection in rural low-computing scenarios. Results Experimental results show that the proposed framework achieves strong overall performance, including 700 ms transmission latency for 50 KB data, a storage cost of 364 yuan/month for 10 GB data, 97.2% accuracy for 10% missing-value completion, 98.2% accuracy for 3% malicious tampering warning, and 390 transactions per second under 700 concurrent transactions. Discussion The proposed framework effectively balances real-time performance, trust, and cost, and provides a practical traceability paradigm for agricultural supply chains. Future work will further improve multi-product adaptability and cross-domain privacy protection.