Blockchain-Based End-to-End Agri-Food Traceability Leveraging Computer Vision, IoT, and Wireless Communication

Blockchain has emerged as a promising technology for agri food traceability due to its ability to pro- vide immutable and transparent records across multiple stakeholders. However, the trustworthiness of a blockchain based traceability system fundamentally depends on the quality, continuity, and physical relevance of the data that are recorded on chain. This thesis addresses blockchain traceability not as a standalone ledger problem, but as an end to end system challenge that spans physical observation, local intelligence, wireless communication, and secure recording. The research investigates how reliable traceability evidence can be generated, validated, and delivered under realistic agricultural conditions before being anchored on a blockchain. Computer vision techniques are first explored as a means to transform physical product characteristics into interpretable and verifiable digital evidence. Using saffron as a primary case study, the thesis demonstrates how visually derived fea- tures can support quality assessment and fraud detection, and how morphology aware descriptors enable the identification of subtle physical manipulation that may evade conventional inspection. These visual outputs are designed to be compatible with blockchain level validation, allowing the ledger to enforce evidence based admission rules rather than passively storing declared attributes. To ensure temporal continuity of traceability records, the thesis further investigates embedded IoT sens- ing architectures that operate under energy and deployment constraints. By combining context aware sensing, adaptive operation, and edge level decision logic, the proposed systems support sustained data generation that aligns with the requirements of long term blockchain traceability. This design highlights the dependency between device level reliability and the completeness of immutable records. The role of wireless communication is examined as a critical prerequisite for blockchain traceability. Through simulation and field measurements in olive orchard environments, the thesis shows that conven- tional distance based propagation models are insufficient to explain link behavior in structured agricultural settings. A direction and geometry aware path loss model is therefore proposed and validated, providing improved predictability for mobile and static communication scenarios. These results demonstrate that communication design directly influences the timeliness and integrity of data delivered to the blockchain. Overall, the thesis proposes an integrated framework in which blockchain traceability is achieved through the coordinated design of evidence generation, validation, delivery, and recording. By grounding immutable digital records in physically verifiable and continuously delivered observations, the proposed approach strengthens the practical credibility of blockchain based traceability systems in agri food supply chains.