Researchers from the University of South Australia recently developed a deep learning system that uses drones to detect areas in agricultural land that require additional irrigation or fertilizers.
The system allows farmers to precisely plan how much water and nutrients they will need on a given day. The method also has the potential to drastically improve crop health, moisture, and nutrient content, making it easier for farmers to focus on the big picture needs of a farm.
“Drones enable farmers to move from traditional farming practices to precision farming, increasing their ability to accurately nurture crops across different sectors, at a reduced cost,” Dr. Zohaib Khan a lead researcher of the study said. “Until now, the drones required an expensive multispectral camera to scan agricultural land and indicate where there is a need for additional irrigation or application of fertilizer to selected crop segments.”
Using NVIDIA TITAN X GPUs, with the cuDNN-accelerated Caffe deep learning framework, the researchers trained their neural network on thousands of frames captured with an RGB camera, the standard on most consumer cameras, mounted on a drone. The multispectral camera was also used to compare the RGB camera’s results.
“The main idea is to learn the spatio-spectral relationships between information in RGB images of vegetation and their corresponding VI values,” the researchers stated in their research paper. “This is achieved by leveraging a deep neural network (DNN) to model the non-linear relationship between an RGB image and its vegetation index.”
The results of the images obtained with the drone are highly correlated with ground measurements of the vegetation index, the team said.
The proposed methodology can also be extended to high-resolution RGB cameras, the researchers said.
“The results prove that RGB images contain sufficient information for vegetation index estimation, the researchers said. “It demonstrates that low-cost vegetation index measurement is possible with standard RGB cameras.”