P.PSH.1263 - Artificial Intelligence Cattle Recognition Pilot
Integrity Systems Company has a vision of a fully automated supply chain, and this project aligns with its strategic target of real-time traceability.
| Project start date: | 14 July 2020 |
| Project end date: | 20 December 2022 |
| Publication date: | 06 November 2023 |
| Project status: | Completed |
| Livestock species: | Grass-fed Cattle |
| Relevant regions: | National |
|
Download Report
(0.4 MB)
|
|
Summary
To maintain competitive advantage, the red meat industry must pursue and invest in new technologies and approaches to integrity that address current and future customer requirements, maximise value and improve uptake across the industry. The aim of the project was to conduct a proof-of-concept pilot using Artificial Intelligence (AI), and video images and advanced computer software algorithms to enable cattle recognition and automatically trace individual animals without the need for National Livestock Identification Scheme (NLIS) tags.
Objectives
1. A database of at least 50 individual cattle on three selected properties will be established such that the AI system can recognise each animal and link it to its on-farm grower information records. Cattle on-farm movements will be used to track animals to test and demonstrate the system accuracy. Suitable cameras will be installed on the farm to track and identify animals.
2. Cameras will be installed at the sale yard used by these properties, and the system will identify the cattle individually as each enters the sale yard and display the growth/provenance information from the property records entered.
3. A presentation and video demonstration covering the on farm and sale yard trials will be developed.
4. The project will consider the scalability of the technology across a full commercial cattle supply chain in Australia (birth to processing of cattle).
5. The project will produce a preliminary commercial viability statement for users of the system (producers, sale yards, processors).
6. The project will produce a preliminary statement on any changes to work methods and the practicality of the system for users (producers, sale yards, processors), e.g. the time it takes to scan and register cattle, time taken to read and identify cattle at unload points, particularly whether unreasonable delays will be experienced.
Key findings
Video of cattle was collected in a feedlot over several months with the camera positioned over the water trough. Lumachain have learnt from this pilot, video-based cattle identification requires a significant number of frontal face images for registration and recognition. The collection method provided less than 5% of useful frames. While high accuracy can be achieved with sufficient frontal images, the true challenge is to have cattle face a camera to collect sufficient images for registration or identification in a reasonable timeframe. While this can be done by restraining cattle in a crush this is not a practical replacement for NLIS tags. To be successful a method must be found to obtain frontal face images of cattle reliably in a feedlot, at saleyards, on arrival at abattoirs and eventually in the field.
This project has achieved confidence that cattle recognition is possible with sufficient front-face images. The practical problem to be solved to be an NLIS replacement is how to achieve cattle frontal facial images reliably without intervention (e.g., crush).
Benefits to industry
This project is the first part of providing a complete solution that provides animal life traceability up to entry to primary processor, thus completing the supply chain ‘paddock to plate’ traceability.
Future research
This pilot has given confidence that accurate recognition is possible with sufficient facial images. However, this project was terminated due to a number of unexpected challenges around the facial recognition technology working in a commercial setting, and that the project required to be rescoped with additional partners to develop the video-based cattle recognition technology.
More information
| Project manager: | Verity Suttor |
| Contact email: | reports@mla.com.au |
