PNIPU: Technology for detecting microplastics in soil, water and air

The main articles are:

• Neural networks (neural networks)
• Video analytics (terms, applications, technologies)
• Computer vision (machine vision)

2024: Announcement of the way to determine microplastics in nature

On July 1, 2024, representatives of the Perm Polytechnic reported that our way of determining microplastics in nature.

source = PNIPU

As reported, according to the State Report for 2022, the volume of municipal solid waste (MSW) in Russia amounted to almost 46 million tons, of which 5 million tons are plastic garbage. At the same time, annually only 14-18% of its total is collected for reuse. Everything else is sent for burial. Microplastic particles are found in all areas of the natural environment, as well as in living organisms, which worsens the ecological situation on the planet. To take timely action to protect the environment, around the world are looking for ways to quickly and accurately find plastic particles in soil, water and air. PNIPU scientists have developed technology for determining microplastics using neural networks and machine vision.

The study was published in the materials of the All-Russian Scientific and Practical Conference "Chemistry. Ecology. Urban "2024. The work was carried out within the framework of the Sirius program. Summer. " Polymer materials decompose for an average of 400 to 700 years. Under the influence of natural factors, for example, direct ultraviolet radiation, they decompose into microplastic particles less than 5 mm in size and are embedded in complex media, mixing with them. It is found in water, soil and some types of food, such as marine fish or plants.

Microplastics take on various forms and have a heterogeneous composition, so determining its number and properties takes a long time. For June 2024, all samples are studied manually using filtration, microscopes, spectral analysis and some physicochemical methods. Such a process is very laborious.

Scientists at the Perm Polytechnic University have developed a way to determine microplastic particles in components of the natural environment using computer vision and neural networks. They trained the neural network to isolate and determine the type of microplastics. Machine learning methods optimize the accuracy of results and the speed of sample processing several times. This will reduce the time and cost associated with errors.

To conduct research, we used artificially prepared samples by crushing several types of plastic: polyethylene terephthalate, polypropylene, low-density polyethylene. The pre-polymer waste was washed, ground and sieved through a sieve with a mesh size of 1 mm. The plastic was then mixed with sand to simulate environmental conditions. A 40-fold magnification microscopy method was used to determine the particles. So we collected an array of a training sample of 100 images, and applied it to training the neural network. shared Kirill Aristov, Master of the Department of Environmental Protection, PNIPU

The resulting data set was divided into three samples: 89% of the images were used for neural network training, 6% for validation, an intermediate check is performed on its basis, and 5% for the test that is needed for final verification.

Effective learning requires a lot of repetition, otherwise the neural network works inaccurately. Therefore, training is carried out in several cycles. The more of them, the better trained the neural network. We produced training for 30 cycles. According to the results, the average recognition accuracy of microplastics was 82.63%. shared Rostislav Kokoulin, Master of the Department of Automation and Telemechanics, PNIPU

The use of computer vision and neural networks are promising and promising methods in solving many environmental problems, including the identification of pollutants in environmental objects. Our research in this area will continue, and we hope to get decent results. supplemented by Natalia Slyusar, Doctor of Technical Sciences, Professor of the Department of Environmental Protection, PNIPU

Scientists of the Perm Polytechnic University have developed a technology for determining microplastics using a neural network. It automates the process of detecting and classifying plastic particles, optimizes information processing and allows you to control the state of the environment.

See also

• Plastic waste