Assessing the performance of recognized ML approaches on remote sensing data for sowing progress detection in Kazakhstan
Alfarabi Imashev (a*,b), Nurali Khamitov (b)

a) Nazarbayev University, Astana 010000, Kazakhstan
* alfarabi.imashev[at]nu.edu.kz

b) LLP SkyTerra, Astana 010000, Kazakhstan n.khamitov[at]skyterra.ai

Abstract

Remote sensing fundamentally entails the acquisition of data
about the Earth^s surface using satellites, drones, or sensors placed on
airplanes, without direct interaction with the terrain. Remote sensing has revolutionized industrial agriculture, providing farmers and researchers with effective tools to observe and manage crops with greater effectiveness and sustainability.

This technology enables monitoring of vast agricultural areas, assisting
in the evaluation of crop vitality, soil moisture levels, insect invasions,
and nutritional deficits. Several main applications encompass:

- Multispectral and thermal imaging, when used for crop identification
and monitoring, can identify plant stress before it becomes visible to the
unaided eye.

- Irrigation management: sensors optimize water use by detecting arid
areas.

- Remote sensing enhances the accuracy of agricultural yield estimations
and predictions via the analysis of vegetation indices, including NDVI,
GNDVI, EVI, and the Canola Index (EAYI), among others.

- Early identification of pathogens and pests facilitates the implemen-
tation of targeted measures, which minimizes chemical use.

Recent advances, including the use of machine learning and cloud computing, have made remote sensing more accessible and accurate than ever before. It is particularly advantageous in precision agriculture, where data-informed
choices can enhance yield while reducing environmental impact.

This paper delineates the second iteration of testing machine learning
approaches to detect the tentative sowing process beginning date (with
the admissible margin of error of +/- 2 days) and to monitor subsequent
sowing progress in Kazakhstan, which was requested by the
Kazakstani National Space Agency to explore the possibility of using ML
solutions for monitoring tasks.

Keywords: Remote sensing, Satellite Imagery, Machine Learning, Computer Vision

Topic: Topic B: Applications of Remote Sensing