Strawberry plant wetness detection using computer vision and deep learning

Abstract

Botrytis fruit rot and anthracnose are fungal diseases of strawberry. These diseases are a significant contributor to yield losses, requiring farmers to use fungicides frequently to prevent them. The proliferation of botrytis and anthracnose is directly linked to the duration of the presence of free water on the plant canopy, which is generally defined as leaf wetness duration (LWD). LWD is an important measure in determining the risk for these diseases to develop in the strawberry crop. By accurately measuring LWD, the risk of disease can be calculated more accurately, and specific fungicide application recommendations can be given to the farmers. This reduces the frequency with which fungicide is applied and ultimately reduces costs for farmers. There is no standard method to detect leaf wetness, but leaf wetness sensors are widely used for that purpose. These wetness sensors are difficult to calibrate and not very accurate, which reduces their reliability. The objective of this study was to find a better alternative to the commonly used leaf wetness sensors. This study implemented color and thermal imaging-based approaches as a solution to the problem of leaf wetness detection in strawberry plants. The proposed method used deep learning and computer vision techniques to detect leaf wetness from color and thermal images. The deep learning model was highly accurate in detecting wetness when compared with the visual observation of the images. It was also found that leaf wetness could be detected with a high degree of accuracy using deep learning with color images. In the future, using the findings of this study, a portable device can be developed to replace the commonly used wetness sensor with a more reliable imaging-based device.