Ecology, Environment and Conservation Paper

Vol 27, May Suppl. Issue, 2021; Page No.(20-34)


Glen Bennet Hermon and Durgansh Sharma


As technological advances are made to sustain the wildlife in our environment there is a requirement of robust non-invasive techniques equipped to distinctly identify individual animals within a species. Identification of individuals from species and tracking of the identified individuals have been conducted with mostly invasive techniques in the past. However, recently multiple methods for non-invasive identification of individuals in the species have emerged, using computer vision algorithms. In addition, the time complexity and space complexity of the technological approaches are far better than the previously used manual approach for individual identification. There exist a wide range of differences among these techniques, based on the patterns in consideration and the approaches used. This amounts to a large collection of techniques that have value and scope in the subject of pattern recognition and thus demand for a comparative study of these techniques, with a discussion on their accuracy, ease of use and their adaptability in various scenarios. This work reviews the different pattern matching techniques among a plethora of algorithms used to identify animal individuals, within various animal species along with their complexity, categorizing them based on the type of pattern used for recognition. The emphasis is towards the insights of computational techniques for various image-based animal biometrics with the intention to automate these processes. It is observed that mathematically modelled identification techniques with algorithmically specialized normalization techniques are the most efficient identification techniques in a broad range of scenarios. This work provides a foundation, so that better algorithms and techniques can be modelled with respect to the existing ones for identification of individuals in a specific species.