Sokovikova N. Computational methods of digital image processing for detection and evaluation of parameters of objects, trailed by own movement

The object of research is the image objects with a non-zero rate of apparent motion. The purpose of research is to improve the accuracy of estimation of the parameters of objects with circular and blurred images due to the development of new and refinement of well-known computing methods of the digital image processing. The methods of research are methods of the theory of computational methods, mathematical statistics theory, optimization theory, which allowed to develop: computational method for estimating the parameters of the image of celestial objects on the CCD frames by using the subpixel model of the object image; computational method for estimating the parameters of the image of celestial objects on the CCD frames, according to the criterion of a minimum sum of squared deviations, when describing the sub-pixel digital image model; computational method of maximum likelihood estimation of the position of objects with image trailed by its own movement; theory of digital image processing, which allowed to improve method of pre-allocating images of celestial objects with image trailed by its own movement; theory of statistical modeling and processing of natural data, confirmed the validity of the theoretical results. Theoretical and practical results consist of the following. The computational methods for processing of digital images, developed in the dissertation, allow increasing the probability of detecting objects at a high rate of apparent motion and improve the accuracy of estimation of the position of objects with a circular image versus to traditional methods. Scientific novelty of the results lies in the fact that first proposed a three-step computational method for detecting images of objects, trailed by own movement. Its use to maximize the probability of detecting objects at a high rate of apparent motion. Improved iterative method for estimating the position of an object with a circular image on a digital CCD-frame based on the sub-pixel model. The method allows reducing the average deviation of the object position estimation with a circular image in comparison with traditional methods, which are based on an approximation of the object image by Gaussian model. Improved computational method for estimating the position of the object with the digital image, trailed by its own motion, based on the sub-pixel image model objects. Using this method allows to expand the limits of reliable measurements of an image with low signal to noise ratio. The thesis work was carried out in accordance with the plan of research work of the Department of electronic computers of Kharkov National University of Radio Electronics in the framework of research work №273 "Neuro-fuzzy systems for the current clustering and classification of the data sequences in terms of their curvature missing and anomalous observations "(№SR 0113U000361, 2013-2015) section №273-2" Adaptive methods and models of data classification and time series prediction in terms of their curvature missing and anomalous observations based on artificial immune systems". The developed computational methods of digital image processing implemented in the educational process in the Kharkov National University of Radio Electronics (act of 01.06.2015), The observatory ISON-NM (code in the Minor Planet Center - H15) (act of 10.03.2014) and software Ukrainian virtual Observatory for processing astronomical CCD frames astronomical surveys (act of 16.04.2015). The developed computational methods can be used in digital image processing software for pre-selection and evaluation of the position of objects with circular images and images, trailed by its own movement, for example in the programs of the automated detection of small bodies of the solar system on a series of digital frames.