MODEL FOR PROCESSING IMAGES OF ONLINE SOCIAL NETWORKS USED TO RECOGNIZE POLITICAL EXTREMISM

Abstract

The scientific research is devoted to solving the important scientific and practical problem of recognizing calls for political extremism in online social networks, which today, due to their high popularity, are one of the main ways of spreading such calls. It is shown that modern means of detecting calls for political extremism in online social networks are mainly focused on the semantic analysis of text messages contained in them. At the same time, in modern online social networks, graphic resources have become widespread, which provide ample opportunities for the implementation of such calls. The possibility of detecting destructive content in images and video materials using neural network analysis is considered. The possibility of increasing the efficiency of neural network recognition has been determined due to the developed image pre-processing model, which makes it possible to adjust the brightness and contrast of images, as well as eliminate typical interference during video recording. The originality of the model lies in the use of a wavelet transform apparatus for filtering typical noise, as well as in the developed mathematical apparatus for adaptive contrast correction based on the local contrast of the neighborhood. It is shows that the use of the developed model for pre-processing images makes it possible to increase the accuracy of neural network recognition of calls for extremism in images and videos posted on online social networks by approximately 12 percent. It is advisable to correlate the paths for further research with the development of a neural network model adapted to the wide variation in the sizes of images and videos in online social networks.