AltSU graduate has developed a system for recognizing birds from photos and videos using artificial intelligence

Alexander Polovinkin, a graduate of the Institute of Digital Technologies, Electronics and Physics of Altai State University, has developed a program for threshold recognition of birds from video and photo images using artificial intelligence. This is the topic of his thesis, which can be useful to airports, factories, agricultural enterprises and private households for which birds are unwanted guests.

Alexander Polovinkin notes that, according to the Federal Air Transport Agency, in Russia in 2017, for every 100,000 takeoffs and landings, there were 3.7 bird strikes. An example of such an emergency situation occurred on August 15, 2019: when taking off from Zhukovsky, Ural Airlines Airbus 321 collided with a flock of seagulls, which clogged both engines, which led to their shutdown. Pilots Damir Yusupov and Georgy Murzin landed a plane with idle engines in a cornfield.

A lot of money is spent on bird control, while not all methods are effective. The easiest way is visual scaring. These can be stuffed animals imitating a person or a bird of prey. Another approach is to install spikes in places where birds most often sit - on roof ridges, monuments, etc. They also try on such inhumane methods as poisoning. Various acoustic and ultrasonic repellers are effective, which broadcast either noise or socially significant sounds, for example, the distress cries of a given bird species or the cries of predators.

The most popular among repellers are gas cannons that produce a loud explosion, simulating a shot that frightens birds. Alexander took it as a basis in his work.

“I conducted the development at the request of the management of the Barnaul brewery. During the unloading of wagons with raw materials for the production of beverages, birds eat up grain and pollute the territory with waste. To combat them, it is necessary to install a bird scaring system. Now there is a propane gun on the territory of the plant, which, due to loud pops, imitates the sound of a shot from a firearm, but at the same time it frightens not only birds, but also employees, since the shots occur at random times. My version involves adding control of this gun – using a neural network. We have some kind of intermediate link that decides at what moment to shoot. With the help of cameras, the program counts the number of birds, and if there are more of them than set by the user, we fire the cannon.”

A big plus of the project is that its implementation does not require any special equipment – ordinary cameras and a computer are enough, the main thing is the software developed by Alexander Polovinkin. Instead of a gun, you can use any other method of scaring off birds.