Use of convolutional neural networks for segmenting images of roads from satellite

The article developed a technique for using convolutional neural networks for automatic segmentation of roads in images obtained from satellites with a synthesized aperture. The analysis of the subject area and the relevance of this study. The development of a neural network based on U-net was carri...

Full description

Saved in:
Bibliographic Details
Published inIOP conference series. Materials Science and Engineering Vol. 971; no. 5; pp. 52048 - 52052
Main Authors Seliverstov, Ya A, Seliverstov, S A, Naryshkin, R S, Kripak, M N
Format Journal Article
LanguageEnglish
Published IOP Publishing 01.11.2020
Online AccessGet full text

Cover

More Information
Summary:The article developed a technique for using convolutional neural networks for automatic segmentation of roads in images obtained from satellites with a synthesized aperture. The analysis of the subject area and the relevance of this study. The development of a neural network based on U-net was carried out in Python 3x using the libraries TensorFlow, TensorBoard, Pandas, Numpy, Scipy, Matplotlib, Sklearn. The neural network was trained on a training sample of 1200 images prepared by hand marking. The accuracy of the developed model when testing on prepared samples was 68%. According to the results of the study, conclusions were drawn and prospects for further functional development of the developed tools were determined.
ISSN:1757-8981
1757-899X
DOI:10.1088/1757-899X/971/5/052048