ABSTRACTS OF ARTICLES OF THE JOURNAL "INFORMATION TECHNOLOGIES".
No. 12. Vol. 24. 2018

DOI: 10.17587/it.24.805-812

A. G. Zlobina, PhD, Researcher, e-mail: ELF54@yandex.ru, I. V. Zhurbin, D. Sc., Chief Researcher, e-mail: zhurbin@udm.ru, Udmurt Federal Research Center of Ural Branch of the Russian Academy of Sciences, Ijevsk, Russian Federation

Comparative Analysis of Segmentation Algorithms of Electric Profiling Data for Restoration of Boundary of Object

In this article assessment of applicability of methods of segmentation for the restoration of the boundary line for the artifacts (underground engineering structures and pipelines, industrial and domestic disposal sites, objects of historical and cultural heritage, etc.) based on geophysical data — the data of areal electric profiling — is carried out. Reconstruction of the shape and geometric parameters allows to correlate the detected local anomalies with search objects of different types. This significantly clarifies the interpretation of the results of the geophysical survey of the territories that have been subjected to human and man-made impact. Methods of classification are applied to the arrays of the apparent resistivity data with the subsequent segmentation of the resistivity distribution map.
The assessment of efficiency of methods of segmentation was carried out on the basis of computer modeling and a natural experiment (geophysical and archaeological studies at the Kushmanskoye (Uchkakar) fortified settlement dating to the 9th—13th centuries). The assessment of quality of segmentation was carried out by criteria of Hausdorff's distance and pixel distance error. Comparative analysis of the results of applying different approaches to the segmentation of soil resistivity distribution maps showed that the algorithms fuzzy c-means and Gustafson-Kessel of fuzzy clustering are the most effective from the point of view of the restoration of the boundary line of object on local anomaly.
Keywords: electric profiling, artifacts, local anomaly, segmentation, boundary line of object, fuzzy c-means algorithm, Gustafson—Kessel algorithm, Hausdorff's distance, pixel distance error, modeling, natural experiment.

To the contents