No abstract

The object of this study is the process of segmentation of images from unmanned aerial vehicles. It was established that segmentation methods based on k-means and a genetic algorithm work qualitatively on images from space observation systems. It is proposed to use segmentation methods based on k-means and a genetic algorithm for segmenting images from unmanned aerial vehicles. The main stages of image segmentation methods based on k-means and genetic algorithm have been determined. An experimental study of segmentation of images from unmanned aerial vehicles was carried out. Unlike known ones, image segmentation by a k-means-based method that successfully works on images from space surveillance systems cannot be directly applied to image segmentation from unmanned aerial vehicles. Unlike known ones, image segmentation by a method based on a genetic algorithm that successfully works on images from space surveillance systems also cannot be directly applied to image segmentation from unmanned aerial vehicles. The quality of segmentation of images from unmanned aerial vehicles by methods based on k-means and a genetic algorithm was assessed. It was established that: – the average level of first-kind errors is 70 % and 51 % when segmenting an image from an unmanned aerial vehicle using methods based on k-means and a genetic algorithm, respectively; – average level of second-kind errors is 61 % and 43 % when segmenting an image from an unmanned aerial vehicle using methods based on k-means and a genetic algorithm, respectively. It was concluded that further research must be carried out to develop methods for segmenting images from unmanned aerial vehicles.

The object of this study is the process of segmentation of images acquired from space optoelectronic surveillance systems. The method to segment images from space optoelectronic surveillance systems based on the Sine-Cosine algorithm involves determining the threshold level; unlike the known ones, the following is carried out in it: – preliminary selection of red-green-blue color space brightness channels in the original image; – calculation of the maximum distance of movement of agents in the image in each brightness channel; – calculation of the values that determine the movement of agents in the image in each brightness channel; – determining the position of agents in the image using trigonometric functions of the sine and cosine in each brightness channel. An experimental study into segmenting images acquired from space optoelectronic surveillance systems based on the Sine-Cosine algorithm was carried out. It was found that the improved method of image segmentation based on the Sine-Cosine algorithm makes it possible to segment the images. In this case, objects of interest, snow-covered objects of interest, background objects, and undefined areas of the image (anomalous areas) are identified. The quality of image segmentation was assessed using the Sine-Cosine algorithm-based method. It was found that the improved segmentation method based on the Sine-Cosine algorithm reduces the segmentation error of the first kind by an average of 21 % and the segmentation error of the first kind by an average of 17 %. Methods of image segmentation can be implemented in software and hardware systems that process images acquired from space optoelectronic surveillance systems. Further studies may involve comparing the quality of segmentation by the method based on the Sine-Cosine algorithm with segmentation methods based on evolutionary algorithms (for example, genetic ones).

This paper contains formal problem definition of predicting unfavorable airborne events during flight. Restrictions and assumptions are put into the prognosis method of unfavorable airborne events during flight. Mathematical apparatus used to build prognosis method is suggested. As a basic mathematical apparatus it is suggested to use, recurrent neural networks (RNN) based on LSTM modules and convolutional neural networks (CNN). Analysis of these neural networks has shown that RNN based on LSTM modules are mostly effective when analyzing structured text, such as report of investigation of airborne accidents. In its turn, CNN are effective when analyzing unstructured text, such as text messages about the flight situation based on the information from external sources. Prognosis method of unfavorable airborne events during flight based on convolutional and recurrent neural networks is developed. In case of solving the task of prediction of unfavorable airborne events during flight RNN are used for initial setup of the Embedding layer of the structured training data in the process of hybrid neural network training. CNN are used during the direct operation of hybrid neural network model of prediction of unfavorable airborne events during flight. K e ywor d s : deep neural network; convolutional neural network; recurrent neural network; prognosis; unfavorable airborne event; hyperparameter; accuracy factor.

Проведено аналіз відомих основних методів сегментування при тематичній обробці видових зображень. Встановлено, що не існує загального підходу до класифікації методів сегментування. Розглянуто наступні групи методів: методи визначення порогів, методи, що засновані на кластеризації, текстурні методи, методи виділення контурів, методи зміни областей. Досліджено їх основні недоліки та переваги. Проведений аналіз визначив ряд проблемних питань, які є напрямком подальших досліджень щодо підвищення якості обробки видових зображень.