Artificial intelligence-enabled livestock recognition in transport

Livestock is transported to various locations over their lifetime. While conventional identification methods such as hot iron branding are against animal welfare, the current systems based on electronic ear tags are susceptible to loss, failure, and guilty replacing over transport. An identification system is required to re-identify livestock in incidents when electronic ear tags are lost, removed, or destroyed. This study proposes an Artificial Intelligence (AI)-enabled system to re-identify cattle’s Radio Frequency Identification (RFID) based on facial recognition. A dataset of 2500 cattle images was prepared and tested. It automatically detects cattle’s faces using the You Only Look Once (YOLO) algorithm in real-time with 92.87% mean Average Precision (mAP). It relies on Siamese Neural Network (SNN) architecture which has the advantage of transfer learning to recognise cattle faces. By comparing distance similarity, the system can identify the unseen cattle ID after training. It reaches 95.13% accuracy in identifying cattle by providing only one query image and having 20 image samples per cow on the test dataset. The proposed system can augment the current Australian National Livestock Identification System (NLIS) and streamline the livestock supply chain. It can also be easily deployed in transportation systems, tuned to identify other farm animals.