A method for identifying PET bottles using artificial intelligence
Tatiana Krasnova, a student at the Institute of Information Technology and Programming at St. Petersburg State University of Aerospace Instrumentation, applied machine learning algorithms to identify and locate polyethylene terephthalate (PET) bottles in unauthorized landfills in the city and beyond. PET bottles, popular due to their accessibility and convenience, make up a very significant part of all plastic waste, and their improper disposal becomes a source of long-term environmental damage. Therefore, the development of systems that facilitate the detection of debris is a priority task.
As practice has shown, the use of artificial intelligence can significantly speed up the detection and subsequent sorting of waste. The neural network developed by Tatiana Krasnova operates with an average recognition accuracy of 92%. As the author notes, RTAB-Map technology has also confirmed its effectiveness, providing simultaneous localization of an object and photographing it with real-time map construction.
In the future, the concept assumes full automation of the process. The robot scans the area for waste, in parallel, the neural network recognizes waste classes and remembers their coordinates. Then, based on these data, a 3D map of the area is constructed with the pollution elements marked on it. Then the second robot receives a map from the scout and moves to the specified points to collect waste.
The created prototype of a software and hardware solution for the identification and localization of PET bottles in unauthorized landfills has significant potential for implementation in the field of waste management and can be integrated into government programs for the elimination of unauthorized landfills.
