Deep Learning Techniques for Leaf Health Prediction

Abstract

Food security is one of the most important issues discussed worldwide. Furthermore, it becomes more challenging in countries like India, where major population is vegetarian and farmers still follow old conventional farming methods. Plants’ growth is often affected by viral, bacterial diseases. However, experts’ advice on these plant diseases may be costly and time-consuming matter. Recently, computer vision and machine learning are successfully applied to the smart farming. Plant health can be easily monitored, and syndromes can be easily identified by applying machine learning and image processing techniques, over the conventional methods. Leaves are important part of the plant. It generates food for plants using photosynthesis. Hence, damage to leaf may result in reduced food supply to the plant. This results to lesser growth of the plant and lesser flower and fruit bearing capacity. This paper addresses various bacterial and fungal diseases among plants. Impact of each disease on the leaves such as color and shape is also discussed in this paper. This paper studies various deep learning techniques such as convolutional neural network (CNN) model and learning vector quantization (LVQ) algorithm which can be used to distinguish among healthy and disease plants. Difference between healthy and diseased leave was used to train these deep learning classifier. This paper also addresses remedial actions such as recommendation of specific pesticide and its quantity. It was observed that there exists and tradeoff between practical usage of automated system by farmers and accuracy of the system.