Радіоелектронні і комп'ютерні системи (Mar 2019)

COMPUTATIONAL METHOD OF SMOOTHED OBJECTS IMAGES SEGMENTATION ON DIGITAL IMAGES

  • Артем Витальевич Погорелов,
  • Вадим Евгеньевич Саваневич,
  • Вадим Евгеньевич Саваневич

DOI
https://doi.org/10.32620/reks.2017.4.01
Journal volume & issue
Vol. 0, no. 4
pp. 4 – 16

Abstract

Read online

This article describes the computational method for segmenting smoothed objects images on digital images is developed. To reduce computational costs the segmentation is carried out only in areas extracted by the computational method of single objects images segmentation on digital images. Segmentation of smoothed images is carried out sequentially in each selected segment. To reduce computational costs only segments that contain more than one peak are analyzed. Such segments may contain either fragments of smoothed image, or a statistically dependent image of a compact group of objects. In this regard, the decision about the segmented image type is made based by the difference between the segmentation results of both types of object images: smoothed image and an image of a compact group. As a sign of distinguishing between smoothed image and an image of a compact group the ratio between the number of pixels assigned to the image of a compact group and the smoothed image segmented with the start pixel in one peak is used. Images of objects with non-zero visible motion during the exposure time become smoothed with their own motion and have an image whose shape is stretched along the direction of its apparent movement. At the same time, single objects as well as compact groups of objects are immovable during the exposure time. The form of such images of compact groups of objects is often close to circular. To decide the extent of the segmented image it is proposed to use the degree of elongation of its image on a digital frame. As an indication of the degree of elongation of the object image the estimate of the eccentricity value is used in the paper. A study of the developed method of smoothed images segmentation on digital astronomical images was carried out on digital frames obtained using telescopes from 8 astronomical observatories. The developed computational method allowed making a preliminary decision about the type of segmented image (smoothed image or image of a compact group) directly at the segmentation stage. The developed method of segmentation of smoothed images is applied in the software package for automated detection of asteroids and comets CoLiTec. With the use of the developed method, the segmentation of images of comets, asteroids, NEOs, artificial earth satellites is carried out. When processing digital frames obtained without using the daily reference mechanism, the developed method is used to select all objects in the digital image.

Keywords