Precision farming AI solution

  • #AgriTech
  • #Computer Vision
  • #Data analytics
  • #Geo solutions
  • #Machine Learning

About the Client

Leading AgriTech company that develops an end-to-end service for monitoring, and optimizing the health conditions and productivity of trees.

Business Challenge

The client’s primary intent was to give farmers a solution for automatic ripe fruit detection. To do that, he started developing a technology for the early detection of anomalies and diseases at two levels: an individual tree and the entire plantation.

After building the basis for the solution, the client partnered with Quantum to implement more features, fix ones that were working incorrectly, and cover the data science part of the project.

Solution Overview

Quantum extended the client’s in-house AI team responsible for analytics, cloud technologies, and multisensor data operations.

Using sophisticated data science techniques, we’ve enhanced the existing solution with the following features:

  • Image preprocessing.

To provide farmers with identical geo-aligned (accuracy up to 6 cm) images and the ability of delayed site recognition, we align drone imagery using built-in sensors, then stitch and substitute drone and satellite imagery.

  • Orchards decomposition.

Some models recognize each lane of trees and each tree in it. The resulting dataset can be divided into different sets stratified by the number of oranges per image.

  • Growth analysis.

It is helpful to monitor each tree, compare its status with the normal vegetation life cycle, recognize and calculate fruits, or create other measurements easily.

  • Proprietary models.

Our client used proprietary models for decision-making support. They are saving time, reducing costs, and helping control the farm business.

 

Project Description

Analysis and prototyping phase
We provided a model that identifies fruits and tree health from images with a nearly human level of performance.

The solution we made used a neural network to predict crop yields better and, thus, increase the efficiency of farms. Our client can segment fruit trees using the semantic segmentation approach on multispectral drone imagery.

The challenging point was to connect all the data we received from drones once a couple of months. Also, our team highlighted the metrics we needed for analysis, which was an unusual task.

All in all, this system helped our client to achieve better outcomes.

DevOps part
Quantum set up a complete CI/CD pipeline for delivering the solution from the source code to the production environment. This pipeline covered building Docker images for the services, running unit and integration tests, and deploying the services into the Kubernetes cluster.

We also built a solution for monitoring the effectiveness of resource usage in the business application. It significantly optimizes GPU usage of ML-based services.

These implemented functionalities allow to detect low-efficiency services, improve their performance, and reduce operational costs.

Summary: Quantum’s team made a prototype capable of detecting ripe fruits and assessing tree health with nearly a human level of performance. In addition, we improved the accuracy of data analysis and forecasting, thereby increasing farm productivity by more than 40%.

Let's discuss your idea!

image-1

Technological Details

The major part of this project was written in Python.

We used OpenCV to identify trees and work with images.

After getting pictures from drones, combining and analyzing them was essential, so our team used Python libraries made by Google Keras and Tensorflow for treatment graphs.

The Shapely framework helped to manipulate and analyze planar geometric objects.

Python
Python
OpenCV
OpenCV
Keras
Keras
Tensorflow
Tensorflow

Connect with our experts