Initiative work is underway at Novosibirsk State Technical University (NSTU) to create a stand for automatic diamond sorting using unique machine learning algorithms, thanks to which it will be possible to accurately classify natural minerals by color.
Scientists from the Department of Optical Information Technologies, Faculty of Physics and Technology, NSTU-NETI, have proposed a method for classifying diamonds by color, which will become an alternative not only to manual sorting, but also to existing methods of automatic mineral selection and will allow for accurate consideration of their shades. To do this, they are developing an automatic diamond sorting system consisting of three systems: technical vision, feeding and distribution of diamonds by class. The research is carried out in the laboratory of optical spectrometry of NSTU-NETI using unique equipment — a compact multichannel spectrometer "Kolibri-2". This spectrometer was developed by the Institute of Automation and Electrometry of the Siberian Branch of the Russian Academy of Sciences in cooperation with VMK-Optoelectronics LLC and transferred to the Department of Optical Information Technologies.
The Kolibri-2 spectrometer is characterized by high photometric accuracy, and its optical scheme and design are optimized for obtaining high-quality spectra with low background radiation in the range of 190-1100 nanometers. The device is designed to measure the transmission spectra of solids, liquids and gases. The data obtained with its help allows you to adjust the boundaries of the color ranges of diamonds. The laboratory conducts work on determining the color of objects based on their transmission spectra in the framework of special courses "Optical spectral analysis", "Sources and receivers of optical radiation", as well as "Color Science and color reproduction". Therefore, the ICU department has all the necessary competencies to conduct research in this field.
According to Marina Zavyalova, Head of the ICU Department, Candidate of Technical Sciences, artificial intelligence plays a key role in the development: it will analyze data from the vision system. "AI will be trained on large amounts of information about colors, which will allow it to improve sorting accuracy and adapt to new conditions," said Marina Zavialova.
The technical principle of the sorting stand is that diamonds will pass through a camera-based vision system that will capture their color and then compare this data with the specified parameters to determine the color category. After the analysis, the diamonds will be automatically distributed to the appropriate containers.
"The uniqueness of the development lies in hyperspectral imaging in a wide spectral range. In addition, the new sorting method involves advanced machine learning algorithms that are able to analyze even minimal differences in shades. This will make it possible to more accurately classify diamonds into ten color categories, taking into account the small color variations that are difficult to perform during manual sorting," Marina Zavialova emphasized.
The scientist added that the stand would allow diamonds to be sorted into ten color categories at a rate of five pieces per second and with a sorting accuracy of up to 98%, which would greatly simplify the work of diamond mining companies. Moreover, the development will allow sorting not only natural diamonds, but also other objects and materials in many industries, improving process efficiency and product quality.
They will test the stand at the ICU department in a few months, as soon as the work on setting up the equipment is completed.