System for identifying a tiger by its unique skin pattern developed

System for identifying a tiger by its unique skin pattern developed

10 October 2020

The Automated System for Identifying the Amur Tiger (SIAT) by its unique skin pattern developed with the involvement of the Amur Tiger Centre is being tried out on the territory of the tiger’s natural habitat. The system was designed to help identify and count Amur tigers and also to trace their blood relationships and to identify individual habitats using images captured by the camera traps.   


In 2019, the Amur Tiger Centre started to develop an automated system to identify and count Amur tigers using images captured by camera traps. Computer Vision Systems, a company with vast experience when it comes to developing software based on computer vision and neural network algorithms, was signed up to implement the idea.


“The system had to offer several solutions,” said Mikhail Smirnov, Technical Director at Computer Vision Systems and Director of the Video Technology Department at Lanit-Tercom. “First of all, identifying the tiger in the photograph provided by a camera trap and making sure it is the same tiger, since it might be the image of another animal. Second, identifying each tiger in the photograph, in case there are several. Next, carefully outlining the silhouette of each tiger in the photograph against grass, trees or other natural features that serve as camouflage for the tigers for further identification. After completing all of these stages, it must be checked to see if the tiger has been included in any existing database and then the animal must be identified.”


Smirnov said that developing a solution for using camera trap images to identify Amur tigers had proved to be one of the most challenging and exciting projects for the team. 


According to him, to find a complex solution to the job assigned to his team, it was decided to use the convolutional neural network capable of outlining “masks” (the contours of sought objects to assign class to each pixel in a photograph), even if the objects differ in size while there are several objects in a photograph partly overlapping one another. Then this data is processed by the neural network to, finally, produce a list of Amur tigers living in the area, each of which had to be identified by comparing tigers with one another.


“The test accuracy of the system is 97 percent with the detector – a person who decides if there is really an actual tiger in the photograph; and it is almost 100 percent with the identifier – a person who establishes the identity of a specific tiger,” Smirnov explained. “Hopefully, the system we have developed will be found very useful by researchers and specialists involved in Amur tiger studies based on photo monitoring and will allow them to process big data within the shortest possible time.”


Currently, the system is being tried out by specialists who are conducting photo monitoring of Amur tigers living in the federal and regional protected areas in the Primorye Territory, as well as in the Jewish Autonomous Area known for its efforts to carry out the programme for the reintroduction of the Amur tiger along the boundary of its natural habitat. 


“This software is not just a system for identifying tigers but is also a database that helps store huge amount of information on tigers,” says Sergei Aramilev, General Director of the Amur Tiger Centre. “Given there is online access to the system, we regard it as a single system for storing and processing data on Amur tigers in the future. When the trial has been completed, we will make sure that the specially authorised agencies of all constituent entities on the territory of the tiger natural habitat, as well as Ministry of Natural Resources and Environment offices have access to the system. We need this to create a complete single database to provide information on Amur tigers living in the Russian part of their natural habitat.”