Kovalenko T. Methods and information technology for the segmentation of textured image areas in air monitoring systems

In the dissertation work the actual scientific task of developing an automated method of segmentation of informative textural regions of images close to the values of color and structure, in order to increase the effectiveness of visual analysis of aerial photographs in air monitoring systems is solved. The analysis of the current state of automation of image processing processes in airborne monitoring systems has been carried out and it has been established that there is a contradiction between the amount of information that can provide modern means of air monitoring and limited possibilities for processing these data. One way of resolving this contradiction is by segmenting the image of smaller areas that may include the desired areas. An automated method of segmentation of informative texture areas of images in air monitoring systems on the basis of the proposed set of features was developed. This method consists of stages of primary and secondary segmentation, which provide a solution to the problem of localization of image areas. The central place of the developed technology is taken by the algorithms for segmentation of the informative texture regions of the image by a given set of signs based on a hierarchical approach, which ensures high reliability of the results. These features are based on the statistical and spectral properties of textures.