RORSE
http://conf.rse.geosmis.ru

ALL-RUSSIA OPEN ANNUAL CONFERENCES ON

CURRENT PROBLEMS IN REMOTE SENSING OF THE EARTH FROM SPACE

Principal physics, methods and techniques for monitoring the environment, potentially dangerous phenomena and objects



рус

Information Technologies in Remote Sensing of the Earth - RORSE 2018

Proceedings of the 16th Conference (November 12-16, 2018, Moscow, Russia)

Snow and Cloud Detection Using Convolutional Neural Network According to the Data Derived from MSU-MR Sensor of the Spacecraft Meteor-M No. 2

Lubov S. Kramareva1, Alexander I. Andreev1, Evgeny V. Simonenko1,2, Egor I. Kholodov1, Julia A. Shamilova1, Alexey A. Sorokin2

  1. Far-Eastern Center of State Research Center for Space Hydrometeorology "Planeta", Khabarovsk, Russia
    alexander.andreev.mail@gmail.com
  2. Shared Facility Center "Data Center of FEB RAS", Khabarovsk, Russia
DOI 10.21046/rorse2018.60
The paper presents a method of forming cloud and snow masks using classifiers based on a convolutional neural network. The input data are the textures of six channels, obtained according to MSU-MR sensor in a resolution of 1 km. A comparison was made the Random Forest algorithm, where separate pixels, NDVI and NDSI indices were used as input data as well as texture features calculated from GLCM matrix. The resulting classifiers were evaluated by calculating f-measure as well as by comparison with the results of manual interpretation by an experienced specialist and comparison with the cloud mask derived from VIIRS sensor.
Keywords: MSU-MR, snow mask, cloud mask, machine learning, convolutional neural network, CNN
References:

Download pdf

Section 2. Methods and algorithms for processing remote monitoring data

60-67