Visual sensor network based early onset disease detection for strawberry plants

The ever increasing use of plant protection chemicals (PPCs) has been on the constant rise as the agriculture industry tries to keep up with growing demand. Excessive usage of PPCs leads to smaller profit margins for farmers as well as damage to ecosystems. An internet of things based visual sensor network was developed to feed data into a neural network classifier which would detect the early onset of plant disease. The sensor network was deployed at a farm owned and run by Wilkin & Sons, a soft fruit grower based in Essex, UK. A prototype convolutional neural network was developed with the purpose of classifying 3 types of images; healthy plants, powdery mildew affected plants and leaf scorch affected plants. The classifier was able to reach an accuracy of 95.48 % for late stage disease detection through images alone.