MTUSI Method of tissue sorting in recognition systems Textile control

2025: Introduction of Tissue Parsing Method in Recognition Systems

MTUSI has proposed a method for sorting tissues in recognition systems. The university announced this on April 14, 2025.

In textile manufacturing, product quality control is central because defects such as pattern mismatch, loose threads and color deviations can significantly affect the performance of the final product.

Until recently, most tissue quality control operations were performed manually, which revealed only about 70% of possible defects. Technologies offer new solutions - automated quality control systems based on computer vision - that are already being used in the textile, woodworking and chemical industries.

Automatic quality control system consists of scanner, recognition system and decision-making module. One of the important tasks of such systems is to determine the change in the structural properties of materials during the stage of detecting scrap. To check the quality of plain fabrics, as well as to identify the characteristic features of their structure, various control methods are used, for example, optical methods. However, existing recognition algorithms often do not allow the detection of defects in materials in real time, especially if their number increases. It is noted that one way to increase the efficiency of such recognition systems is to create methods for preliminary analysis of a textured geometric spatial model (tissue) and improve algorithms for detecting tissue defects in systems with reference models.

MTUSI scientists have proposed a method for forming textured image features, which can be used to build a mathematical model of textile material (fabric) in solving problems of automating the sorting process. The innovation of this method is the creation of standards - samples with which images of tissues are compared to identify defects.

The analysis of tissue quality relative to the reference state makes it possible to distinguish significant features of the tissue structure up to the analysis of the state of single weaves, - said Sergei Aleksandrovich Rozhkov, professor, Doctor of Technical Sciences. "At the same time, in this formulation of the problem, it is advisable to consider the tissue as an object of control as a two-dimensional periodic object."

In the study, an interesting tool was the use of spatial autocorrelation function, which helped scientists analyze the structural elements of tissue.

The development of the model was carried out in the MATLAB environment. When highlighting the contours of the image, the peculiarities of the periodicity of the tissue structure were revealed, and the presence of a correlation maximum made it possible to determine the main frequencies of the structure, - Vyacheslav Igorevich Voronov, associate professor, Ph.D., told about the course of the study. - To improve the quality of images, image pretreatment methods were used: brightness correction, noise elimination and image transformation for more detailed analysis. At the same time, correction of the brightness scale made it possible to eliminate systematic distortions of the original image of additive or multiplicative nature.

During the study, the scientists drew attention to the fact that the correlation not only enhances the main frequencies, but also reveals hidden periodicity in the tissue structure, which greatly complicates the task of determining the main parameters of the model structure. In turn, the use of harmonics associated with uneven tissue structure in convolution made it possible to distinguish latent periodicity in the direction of the tissue base and to distinguish the skew that occurred in the tissue sample. To determine the fundamental frequencies of the tissue model, scientists used the energy spectrum of the image, with the help of which a reference monogarmonic tissue model was built.

The results obtained in the work cover only one of the possible approaches in the quality control of textile materials. Further studies are promising, from the point of view of creating control systems, for the entire spectrum of defects in textile materials (fabrics), including the possibility of predicting their characteristics.

In the future, it is planned to create quality control systems capable of predicting and analyzing in detail defects in textile materials during their production. The improvement of the means of automating the quality control of textile materials allows not only to improve the efficiency of production, but also to improve the quality of textile products, satisfying high standards. industries