Researchers at the Faculty of Computational Mathematics and Cybernetics of Moscow State University have introduced an advanced neural method for detecting and matching key points on iris images. This is reported by the
official website of Moscow State University.
The work, carried out within the department’s specialised seminar on image processing and computer modelling, presents a hybrid system that integrates classical mathematical techniques with modern neural network architectures. At its core is an enhanced version of the model, which extracts features across multiple scales. It combines first- and second-order image derivatives with trainable convolutional filters, while additional Hermite-based convolutions increase the reliability of key-point detection.
To train the system, researchers developed a synthetic dataset that applies random geometric transformations – such as scaling, shifting and rotation – to normalised iris regions. Photometric augmentation, including variations in brightness and contrast, further strengthens the model’s ability to generalise across challenging imaging conditions.
According to the research team, the results confirm the method’s practical potential in addressing real-world biometric identification tasks. The approach underscores the growing relevance of hybrid techniques that merge mathematical rigour with neural adaptability – an area expected to drive the next wave of innovation in biometric technologies.
Photo: ismagilov /
iStock
Самые
актуальные новости стран БРИКС https://tvbrics.com