Information technology of early crop identification by using satellite images

Abstract

The paper deals with the development of information technology for early crop identification in the region of interest by using satellite images. With the early identification being performed in the first half of the growing season, it is characterized by the lack of ground-based and space-based data to configure recognition algorithms. The proposed technology allows one to generate a training dataset by using information from the past years and then use it for crop recognition for the current year. The technology consists of two stages. At the first stage, models of time series, crops and agroseasons are built using data from the past years. At the second stage, data of the current year are processed. First, a suitable model of agro-season, which is close to the development of crops in the current year, is selected by using a tiny set of control fields with known crops. Next, a training dataset is generated based on the characteristics of the agro-season model and the recognition of crops in the current season is performed. The paper assesses the quality of recognition offered by the proposed information technology and the possibility of its application in a particular region. Data collected for the years 2011 -2015 in the Samara Region are used for the experimental research.