Supply chain traceability of live Southern Rock Lobster exports

This project will develop an innovative traceability system for the Australian Southern Rock Lobster (SRL) industry, leveraging cutting-edge computer vision and machine learning to uniquely identify lobsters.

Enabling individual tracking from ocean to plate, the system will enhance supply chain transparency, protects against seafood fraud, and fosters consumer trust in key export markets. It allows for individual-level quality assessments and checks, offering valuable insights to improve handling and logistics while ensuring premium product standards.

This initiative supports a modern, data-driven seafood industry, promoting sustainability, accountability, and premium product standards.

Participants

• Fiordland Lobster Company Limited
• University of Tasmania

Project background

This project will develop a novel tracking and traceability tool for the SRL (Jasus edwardsii) industry to provide enhanced visibility across the supply chain. At its core, the tool will uniquely identify individual lobsters.

Using advanced computer vision and machine learning techniques, this approach will lay the foundation for tracking and traceability from ocean to plate. Individual identification is the fundamental first step in capturing and linking important metadata, such as catch region, size, and condition, which will contribute to product transparency and authenticity throughout the supply chain. Such a system has the potential to provide valuable data for monitoring transport conditions and addressing issues that impact product quality.

Developing a traceability system would offer opportunities for data-driven insights that could help optimise the supply chain. Analysing data generated when lobsters are scanned at key points in transit would help validate supply chain models, ensuring that lobsters are following expected paths and revealing any deviations or undocumented steps.

This capability could help safeguard against product loss or unauthorised reselling while providing new market insights and opportunities by offering visibility into where lobsters are sold after reaching their initial destination.

The need for such a system is underscored by the critical issue of seafood fraud and mislabelling, which costs the Australian seafood industry an estimated AUD 189 million annually.

Establishing the provenance of each individual lobster will result in the system will enhancing visibility throughout the supply chain, protecting legitimate exporters, and fostering consumer trust, ultimately safeguarding Australia’s reputation in key export markets, such as China.

Additionally, further development of the system would improve supply chain and market logistics by enabling efficient management of events such as product recalls (such as during toxic algal blooms). The ability to trace the region of origin and export documentation for affected lobsters would facilitate quicker, safer responses to market disruptions, thereby protecting human health and ensuring that business operations resume promptly.

Ultimately, this approach will ensure the integrity of Australia’s seafood exports and establish a modernised, data-driven SRL industry.

Project objectives

The primary objectives of this project are to:

1. Develop a robust computer vision and machine learning-based pipeline to uniquely identify individual lobsters, enabling the potential for full end-to-end traceability and real-time tracking throughout the entire supply chain—from the point of capture to the consumer
2. Achieve near 100% accuracy in lobster re-identification, ensuring reliable and scalable real-world performance in diverse operational settings.
3. Integrate the identification analysis pipeline into a useable proof-of-concept form, such as a web-based application, with static database to test functionality.
4. Lay the groundwork for a comprehensive traceability and logistics platform that not only ensures the provenance and authenticity of each lobster but also tracks and optimises the entire supply chain.

UPDATE: July 2025

Work has progressed simultaneously on the first two project milestones:

1. Dataset Finalisation; and
2. Image Pre-processing Pipeline.

Image pre-processing is ahead of schedule, with methods for glare reduction and contrast enhancement trialled. A key output has been the development of an oriented bounding box detection model capable of locating both the carapace and individual spines within input images. The current model achieves over 95% detection accuracy, with further training underway.

Detected spines are used to generate binary maps based on their relative size, location, and colouration. These binary maps will form the basis (i.e. the ‘fingerprints’) for individual re-identification in the next project phase.