In a field, a scientist is necessary for detecting disease and finding the most appropriate way of treating it. Those positions cost a lot and for good reason. Scientists need to be trained and get degrees on this field so that they can work on improving crop viability in a field.
A drone with a colour camera was programmed to detect and identify particular plant diseases. Specifically Fusarium wilt of radish, and based on the criticalness of the disease, give a score. We have seen other drones able to spray a field, maybe those technologies can merge and be able to spray a field selectively, eliminating the need for multiple scientists working in a large field or improving their effectiveness with those new tools.
Those drones use GoogleNet for the detection of the disease based on position, size, colour and texture. This is a neural network that can be trained to process and identify visual targets. Based on the images collected by the drones, a score was produced and added to a dataset. This was then processed to assess the criticalness of the disease in each plant.
The system was trained by using 2000 healthy, 2000 light Fusarium wilt and 2000 heavy Fusarium wilt. This took a lot of processing power and a lot of time. The system used included a high end workstation i7 5930K processor, 64GB of ram and a Titan X graphics card. And while this is a bit outdated, those specs are more likely to be used in scientific situations, where reliability is more important and an old and tested system is better than a younger one. But still this shows the potential of this technology, because even with this testing, the result was an accuracy of over 90%.
In the future, high resolution images and modular infrared lens modules will improve accuracy an even allow us to detect multiple diseases with smaller details that characterize them.
Do you like this article? Have anything to add? Please comment or share it with your friends. This is a new section on Qul Mind about technology, we hope you like it.