Evaluation of retinal imaging technology for the biometric identification of bovine animals in Northern Ireland

Animal identification is a major requirement for government agricultural authorities, facilitating registration of animals, recording of authorised animal movements, national herd management, payment of appropriate grants and subsidies and as a vital tool in tracing diseases of public and animal health concern. Most schemes are based on a computer database of ear-tag numbers. A potential limitation of such systems has been their tracking of a device attached to the animal, rather than tracing the animal itself. This becomes problematic when accidental loss or fraudulent switching of tags occurs, as preserving correct identification is difficult. Biometric identification offers considerable advantages since the indices used to construct identifiers are unique, unalterable biological properties. We have evaluated a novel technology in bovine animals as a means of scientifically verifying tag-based identity. This system records the unique retinal vascular pattern at the back of an animal's eye as a means of corroborating ear-tag information. 869 animals were imaged to create a retinal identification. Each of these 1738 retinal patterns were compared computationally and visually against each of the remaining 1737, a total of 1,509,453 comparisons. None of these comparisons yielded an identical retinal pattern, indicative that each is unique within this dataset and that the chances of finding two different eyes with an identical pattern is at least 1 in 1.5 million. A further 2266 images taken from the registered animals at later dates were used for successful identification verification with 2227 (98.3%) being successfully computationally matched to the initial images. The remaining 39 (1.7%) were successfully matched by visual inspection. A simulated tag switch involving alteration of ear-tag number for 115 animals, after their initial imaging, was detected for all participants by subsequent imaging and computational comparison. These data indicate that this system could be deployed as a stand-alone technology for animal identity verification. It also has the potential to improve the performance of ear-tag-based identification systems in cattle and could be deployed in support of identification, registration and movement requirements and as a counter-fraud measure.