Investigating the use of AI-based image identification to monitor wild cryptic lizards: The shore skink, Oligosoma smithi, as a case study.

Pauline Raphanaud, BSc

Master's thesis

Abstract:

Biodiversity conservation increase through the last decades, but is facing massive threats, mainly by habitat modification. Reptiles (Sauropsidae, excluding Aves) are strongly linked to their habitat, and have less movement capacity compared to many other animal taxa such as birds or mammals. Therefore, monitoring reptile populations is a viable method to monitor habitat changes. But reptiles also suffer from of lack of knowledge on their basic ecology, such as movement or even population estimation. This due to less attractive image to the public, and a difficulty of trapping and identifying. Trapping small lizards for Capture-Marking-Recapture surveys requires an investment of time and money, for often poor results, due to trapping rate of the target specie. Using artificial intelligence software to identify individuals by their natural marking is preventing induced harm to the animal by marking and allow researchers to process bigger databases for a reduced time of analyses. In this master’s thesis, we investigated the use of IBEIS, an artificial intelligence software based on image identification, to monitor the Shore skink, Oligosoma smithi, as a case study. We used dorsal pictures of the skinks, among a database of 391 pictures of skinks captured in the pitfall traps of 3 grids in the Tāwharanui Open Sanctuary, with survey sessions were conducted bet November 2006 to May 2008. We identified 10 recapture events with IBEIS, and using SECR 1.4 App, we found out an estimated population of 6864 individuals, with around 820 individual per hectare on our study site. Despite a strong original database, recaptures did not occur frequently, thus we cannot use these estimations as solid proof, but are anyway a first step into understanding the unknown movements of this species. From our results, we think this low recapture rate is not due to IBEIS, but a high density of animal. To have a better understanding of these animals, a survey should be conducted with new technologies we dispose, like camera trapping and AI, following individuals without survey interruptions over year.

Keywords

Monitoring, Recapture, Photo-identification, Lizard, Artificial Intelligence

Language used: English

Date on which the thesis was submitted / produced: 15. 8. 2022

Thesis defence

• Supervisor: doc. Francisco Ceacero Herrador, Ph.D.

Czech University of Life Sciences Prague

Faculty of Tropical AgriSciences

Master programme / specializace:
Wildlife and Livestock Production, Management and Conservation / Wildlife Management and Conservation