File(s) under permanent embargo
Convolutional Neural Networks for individual identification in the Southern Rock Lobster supply chain
journal contribution
posted on 2023-05-20, 15:32 authored by Vo, SA, Joel ScanlanJoel Scanlan, Paul TurnerPaul Turner, Robert OllingtonRobert OllingtonIn most traceability system, product identification is the key to enable the tracking activities along the supply chain to be carried out. Product tagging using barcode and RFID is the most common method for this purpose. However, these practices can be challenging for small businesses due to the cost and time burdens. In addition to this, concern of fraudulent activities such as cloning, substitution and reuse becomes a tangible risk in complicated and poor controlled markets. In approaching Southern Rock Lobster (SRL), this paper proposes an image-based identification solution for individuals using Convolutional Neural Networks (CNNs). This work is built on the prior research of automated lobster grading by the authors, with this next step working towards a low-cost biometric recognition solution for lobster tracking from catch to consumers. In this approach, a Siamese model combined with a contrastive loss function was adopted to distinguish between individual lobsters based on carapace images. These areas are believed to contain individually recognisable features formed by colours and spiny patterns. Preliminary experiments on an image dataset of 200 individual lobsters collected at a lobster processor show the feasibility of using lobster images as an additional factor to provide increased security to the current tag-based tracking systems in use within the SRL supply chain.
History
Publication title
Food ControlVolume
118Article number
107419Number
107419Pagination
1-7ISSN
0956-7135Department/School
Student Life and EnrichmentPublisher
Elsevier Sci LtdPlace of publication
The Boulevard, Langford Lane, Kidlington, Oxford, England, Oxon, Ox5 1GbRights statement
© 2020 Elsevier Ltd. All rights reserved.Repository Status
- Restricted