Biomass Food Crop Prediction
Enhancing Crop Yield Prediction with AI-Driven Drone Imagery Analysis
problem
The client, a company based in Central America specializing in AgriTech, faced the challenge of accurately predicting crop yield before harvest time. This capability is crucial for agricultural businesses to forecast selling and storage capacities effectively, enabling them to optimize crop production and resource allocation. However, the client lacked a solution to estimate the quantity and weight of fruits within a specific area, hindering their ability to manage resources efficiently and impacting their budget planning across multiple fields.
problem
The client, a company based in Central America specializing in AgriTech, faced the challenge of accurately predicting crop yield before harvest time. This capability is crucial for agricultural businesses to forecast selling and storage capacities effectively, enabling them to optimize crop production and resource allocation. However, the client lacked a solution to estimate the quantity and weight of fruits within a specific area, hindering their ability to manage resources efficiently and impacting their budget planning across multiple fields.
solution
To address the client s challenge, we developed a desktop application that leverages drone imagery to estimate crop yield and biomass within targeted agricultural areas. The application incorporates computer vision and artificial intelligence algorithms to analyze drone data, calculate biomass, and predict fruit ripeness, providing the client with valuable insights for biomass management. Open Source Geospatial Foundation (OSGEO) tools, including QGIS and pix4d, were utilized for data processing and visualization, while a diverse technology stack comprising GDAL, NumPy, shapely, geopandas, and laspy supported various aspects of the project.
solution
To address the client s challenge, we developed a desktop application that leverages drone imagery to estimate crop yield and biomass within targeted agricultural areas. The application incorporates computer vision and artificial intelligence algorithms to analyze drone data, calculate biomass, and predict fruit ripeness, providing the client with valuable insights for biomass management. Open Source Geospatial Foundation (OSGEO) tools, including QGIS and pix4d, were utilized for data processing and visualization, while a diverse technology stack comprising GDAL, NumPy, shapely, geopandas, and laspy supported various aspects of the project.
results
The project involved collecting and preprocessing data from the client s agricultural drones, followed by the development of an automatic fruit recognition system. Despite challenges such as artifacts from drone imagery, such as shadows and overlapping images, the team successfully implemented solutions to enhance accuracy. By utilizing digital elevation model (DEM) data and LASer (LAS) data, an ML model was trained to identify plant height and extract data points, enabling accurate biomass classification and prediction. As a result, the client can now effectively differentiate between usable biomass (fruits and hay) and futile biomass, enabling meticulous planning and management of resources across their fields.
results
The project involved collecting and preprocessing data from the client s agricultural drones, followed by the development of an automatic fruit recognition system. Despite challenges such as artifacts from drone imagery, such as shadows and overlapping images, the team successfully implemented solutions to enhance accuracy. By utilizing digital elevation model (DEM) data and LASer (LAS) data, an ML model was trained to identify plant height and extract data points, enabling accurate biomass classification and prediction. As a result, the client can now effectively differentiate between usable biomass (fruits and hay) and futile biomass, enabling meticulous planning and management of resources across their fields.
tech stack
tech stack
