Overview
This project aims to transform the prawn sorting process by introducing hyperspectral imaging (HSI) that distinguishes between soft-shell and hard-shell prawns using their spectral images processed by ML. Current manual methods, which rely on tactile assessment, are subjective, labour-intensive, and inconsistent. By utilising HSI and ML, this research seeks to improve the accuracy, reliability, and efficiency of shell hardness detection.

Objective
The Calibration Phase of this project focuses on developing and validating ML models capable of analysing HSI data to classify prawn shell hardness levels. These models will lay the groundwork for implementing systems in the next phase, ensuring consistent, high-quality products for the market.

Methodology
The project will:

- Rent and deploy a Resonon PIKA L HSI camera, selected based on prior proof-of-concept findings.
- Collect and classify 2,400 prawn samples from the host farm, APF, based on shell hardness and size.
- Generate spectral and morphological data, creating a robust dataset for ML training and validation.
- Develop ML models to correlate spectral signatures with shell hardness levels.
- Validate ML models against independent samples to ensure reliability.
Budget Highlights
The project requests $112,583 from FRDC for salaries, operational expenses, and HSI camera rental. DPI will contribute an additional $35,437 and the host farm will also contribute $6,650 for labour, prawn, packaging and shipping, bringing the total project cost to $154,670.

Team
The project team is led by Dr. Iman Tahmasbian, a Senior Scientist at Queensland DPI with extensive expertise in hyperspectral imaging (HSI) and a strong track record in R&D and high-impact publications. Supported by DPI’s state-of-the-art HSI and food laboratories, the team also includes Dr. Jing Wang and an AgTech data scientist, who are experts in image processing, data analysis, and ML.

Next Steps
Upon successful calibration of the ML models, the project will move to the Implementation Phase, focusing on integrating the developed technology into sorting systems for prawn farms. The implementation phase is not part of this project and will be a future funded project.

This project represents a vital step towards modernising the Australian prawn industry by combining advanced imaging and ML technologies, ensuring a more sustainable, profitable, and high-quality production system.

This program of works supports the trial, implementation, and evaluation of innovative and alternative low-impact harvest technologies (fishing gears) within Queensland’s inshore fisheries (East Coast and Gulf of Carpentaria). The program will be developed in a way that supports an evidence-based approach to developing and trialling sustainable alternative commercial fishing gears, and be run in collaboration with relevant Government agencies and fisheries stakeholders.

Through the additional support of co-investment of $4.5 million by FRDC (in line with this application and the approved funds listed under project 2023-154), the program will be delivered in two stages across a six-year timeframe (up to a total investment of $9 million).

The fishing methods to be trialled as part of the first stage will range from exploring enhancements of existing low-impact gear types, through to trials of innovative harvest technologies. The alternative low-impact harvest technologies will first be trialled in order to demonstrate their triple bottom line credentials. After this, the second stage will support broader implementation and evaluation of commercial application over a three-year period.

The program of works will also explore additional opportunities to enhance the economic value and social profile of the fishery, to ensure that any new harvest technologies align with global best practice standards, product value adding to enhance profitability margins, and improvements in social acceptability.

Wild caught seafood also allows for a diverse mix of species that appeal to a range of consumers and seafood businesses. It is planned that this work will explore potential opportunities for market expansion.

As one of the most prominent species sold in northern Queensland fish shops, Spanish mackerel is iconic to seafood consumers as well as commercial, recreational, and charter fishers. However, a recent assessment of the Queensland East Coast stock raised significant concern about its sustainability, with a spawning biomass estimate below 20% of unfished levels (Tanimoto et al. 2021). Historical analyses indicate a substantial contraction of the spawning aggregations of Spanish mackerel in North Queensland, with those off Cairns being extirpated in the 1990s (Buckley et al. 2017). Previous stock assessments have also shown a trend of declining abundance over the last 20+ years (Welch et al. 2002, Campbell et al. 2012, O'Neill et al. 2018). This historical sustained decline, combined with the recent low biomass estimates, raise substantial uncertainty over the continued viability of the fishery.

This stock assessment result indicates the need to rebuild the stock in line with State and National policy objectives to reduce the number of overfished stocks. The high social and economic importance of Spanish mackerel in Queensland has led to significant public debate around uncertainty in the stock assessment, and the issue being debated in Queensland State Parliament.

As a result, key research gaps need to be addressed to increase the precision of future assessments and reaffirm stakeholder confidence in the stock assessment process. This project aims to address the following research gaps highlighted by the stock assessment: (1) determine the feasibility of using a CKMR method for estimating abundance of Spanish mackerel, (2) quantify cryptic mortality from shark depredation and post release survival, (3) improve catch and effort standardisation by better adjusting for potential hyperstability occurring at spawning aggregations and make the catch rate robust against potential management change, (4) assess the influence of key environmental variables on abundance and recruitment variability. This project will inform more accurate Spanish mackerel stock assessment, thereby increasing stakeholder confidence in line with Outcomes 4 and 5 of the FRDC R&D Plan (‘optimising benefits for the Australian community through fair resource sharing based on evidence-based management’ and ‘transparent decision-making tools that demonstrate best practice in fisheries management’). The project will also directly address the high priority research needs identified by the Queensland Spanish Mackerel Fishery Working Group.