Dudinova O. Methods of intellectual processing of spatial data in geoinformation systems of ecological monitoring

Українська версія

Thesis for the degree of Candidate of Sciences (CSc)

State registration number

0421U103554

Applicant for

Specialization

  • 05.13.06 - Інформаційні технології

23-09-2021

Specialized Academic Board

Д 64.050.07

National Technical University "Kharkiv Polytechnic Institute"

Essay

In dissertation work the solution of the actual scientific and practical task of working out of methods of intellectual processing of spatial data in geoinformation systems of ecological monitoring, which allows to improve the quality of formation of landscape digital images for further analysis of the state of monitoring zones, is proposed. The paper proposes: a method of categorical classification of objects in the tasks of computer analysis of aerial photographs; a method of neural network processing of noisy digital images that may contain distorted fragments, based on the use of a neuroevolutionary model of noise-suppressing auto-encoders; the method of segmentation and allocation of spatial digital images, based on the use of Markov models, which allows to take into account the nature of the area of the analyzed pixel and to set the relationship between the classes of neighboring pixels; the method of neural network processing of noisy cartographic data GIS, which provides implementation of parallel computing structures of pre-filtering of Half-Ton spatial images and noise-proof detection of contours of image objects; Method of correction of color cartographic images in order to improve their quality, which is carried out by means of gamma correction; raster data compression method where combined use of genetic optimization and fractal methods of compression of spatial images is used. The practical results of using the conducted research are algorithms, applied programs and information technology, implementing the methods of intellectual processing and digital images in the GIS of environmental monitoring.

Files

Similar theses