How Drones and Machine Learning Help Farmers

Unmanned aerial vehicles (UAVs), more popularly known as drones, are probably one of the most multipurpose innovations the world has ever seen. Originally developed for military use, drones were rarely applied to other purposes in the past. Today, however, consumer, enterprise, and civil government UAVs vastly outnumber military ones. Civil drone production is forecast to total $73.5 billion worldwide in the next decade, soaring from $2.8 billion in 2017 to $11.8 billion in 2026, with commercial use being the fastest growing civil segment.

Drone technology develops at a staggering speed, promising to affect almost every industry, from logistics, energy, and healthcare to sports, entertainment, and the arts. Drones prove reliable and efficient in a wide range of use cases, including filmmaking, forest fire detection, search and rescue, crowd monitoring, and medical supply delivery to inaccessible regions. Small and nimble drones are extremely beneficial in places where people cannot reach or perform safely, and their full potential has only begun to be explored.

Drones in Agriculture and Farming

Along with construction and communication, agriculture is now among the fastest growing markets for the commercial drone industry. Having to deal with increasingly complex challenges, such as climate change, irrigation problems, and soil quality, just to name a few, more and more growers are turning to innovative solutions and adopting precision agriculture technologies to optimize performance.

Precision agriculture is a farming management concept focused on harnessing geographical, environmental, and biological data to observe, measure, and respond to variability in crops, forecast water and fertilizer demand, schedule resource allocation, and as a result, ensure efficiency, maximize productivity, and improve the quality of agricultural products.

In addition to the multiple tools farmers can use to collect data, agricultural drones represent a new and affordable way to constantly monitor crop conditions by air. Only drones offer farmers such a rich picture of their fields, allowing them to track their crops remotely, combat water shortages and minimize pests, identify any issues in real time and resolve them early on, and improve crop yields significantly.

Auriga Develops Software for Smart Farming

Thus, agricultural drones play a vital role in the future of farming. With this in mind, Auriga participated in a smart agriculture project by developing a drone software utility to monitor the condition of trees in an apple orchard. The idea is that a camera-equipped drone flies over the orchard and tracks changes in the trees’ leaves and the outer bark. Based on this data, a farmer can easily assess the trees’ health, fine-tune irrigation, spot bacterial or fungal infections, and take fast action to protect the trees.

Auriga’s task was to deploy TensorFlow—an open-source software library for dataflow programming, often used for machine learning applications such as neural networks—to a Linux OS-operated computer. At the same time, it was important to provide calculations on the video card to speed up data processing. In other words, Auriga’s team had to install CUDA, cuDNN, Python, and TensorFlow versions compatible with each other and the video card, deploy a ready-made neural network (SSD, a unified framework for object detection based on the MobileNet network), and prepare this entire set of tools for the user.

On a Windows PC station, our engineers labeled images and video for machine learning. The network was already trained to recognize a large number of varied objects, and we continued training it on the apple orchard data. We also enabled augmentation (i.e., increasing the volume of the training sample by duplicating the original data with some artificial variations, such as rotation, shift, and scaling), split the data into training and test datasets, translated the data into the format required by the neural network, and simplified the use of the application.

Along with some other recent projects on machine learning for various industries, including connected cars and healthcare Big Data solutions, Auriga has significantly enriched its expertise in digital transformation.

Elena Baranova, Head of Auriga’s Engineering Department, comments:

Drones, IoT, Big Data, AI, machine learning, and deep learning—technologies are evolving at jet speed, revolutionizing all industries, including agriculture. Modern farming is fueled by the latest innovations and high-tech instruments able to collect, process, and analyze loads of data, turning it into invaluable insights and accurate predictions. With all this data—far more than any single farmer could ever grasp—smart farming becomes possible.