Artificial intelligence is used in agricultural environment monitoring through the integration of mobile robots and aerial imagery. The main goal is to develop automated solutions for crop analysis, with a focus on early disease detection and understanding plant structure.
Modern agriculture requires efficient and scalable methods for monitoring crop health. The use of drones and mobile robots enables rapid data collection from the field, but analyzing this data presents significant challenges.
In this context, AgroBots proposes:
- automation of the agricultural monitoring process
- use of AI for interpreting aerial imagery
- early detection of issues that may affect production
Plant disease detection from aerial imagery
One of the main objectives is identifying leaf-level diseases using images captured by drones.
Challenges
- leaves are often observed from different angles
- perspective-induced distortions occur
- variations in illumination affect visual analysis
Approach
To facilitate the analysis, the following are performed:
- leaf detection from aerial imagery
- correction of their orientation (geometric rectification)
- transforming leaves into a canonical frontal view, similar to laboratory images
This normalization enables more accurate comparison and the application of robust disease detection methods.
Leaf orientation detection
Another key aspect is estimating leaf orientation, which is necessary for correcting perspective.
Contribution
A novel neural architecture has been proposed for detecting object orientation in aerial imagery.
This:
- identifies objects (leaves)
- estimates their direction and orientation
- enables the integration of geometric information into the analysis process
The model is based on the concept of Directional Object Detection, being specifically designed for aerial scenarios.
Impact
By combining:
- drone-captured images
- artificial intelligence methods
- and mobile robots
AgroBots contributes to the development of intelligent systems for:
- automated crop monitoring
- early disease detection
- optimizing agricultural interventions