Rapid assessment of soil salinity using low-altitude unmanned aerial platforms (AP23488745 – RASS 2024-2026)
- General concept of the project
2.1. Introductory part
Soil salinity is a significant factor limiting the development of crop production. Temporal and spatial variability makes it difficult to select optimal solutions for managing agricultural production on soils prone to salinity. Available methods for assessing soil salinity usually have low spatial resolution, are incomplete, expensive and slow. They are not intended for prompt identification of local salinity areas. In order to overcome these limitations, it is proposed to explore the possibilities and methods of rapid assessment of salinity with high spatial resolution based on the use of UAVs and machine learning algorithms. The project is carried out at the intersection of the following scientific areas: artificial intelligence, crop production, remote sensing of the Earth, control of autonomous robotic systems.
2.2. Objective of the project.
Development of a method for rapid assessment of land salinity based on machine learning using unmanned aerial platforms.
2.3. Project objectives.
To achieve the project goal, it is necessary to solve 4 main (1-4) and one auxiliary task (5):
- Development of an unmanned aerial platform.
- Data collection and preparation.
- Development and/or adaptation and tuning of machine learning models.
- Development of a prototype system for mapping surface soil salinity.
- Formation of the final report. Preparation and publication of articles in scientific publications in accordance with the requirements of competition documentation.
As a result of task 1, an unmanned aerial platform will be developed to perform the task of obtaining images of fields for mapping and assessing salinity. As part of task 2, field research will be carried out with the collection of soil samples and obtaining multispectral images of fields using the developed UAP (task 1). The electrical conductivity of soil samples will be measured in laboratory conditions. Based on the data obtained, data sets will be generated for training and tuning machine learning models. The results of task 2 will be used to complete task 3, where the developed data sets will be used for computational experiments to configure and train machine learning algorithms. The results obtained in the form of trained models and program codes will be used to solve task 4, within the framework of which a salinity mapping system will be developed. The results of tasks 1-4 will be used when completing task 5 to generate the final report on the project and for the preparation of scientific publications.
