Intelligent Plant Disease Diagnosis System based on a Multi-modal Approach

Abstract

Plant disease diagnosis systems play a key role in the identification and identification of cases of infection within the plant. Currently, most automatic disease diagnosissystems detect diseases either by analyzing the visual feature of the diseased plant image, where the user query is an image of the disease taken from the field, usually of low quality and also has a complex background. This leads to difficulty in making the appropriate diagnosis. At the same time, the other approach to disease diagnosisis based on a natural language query describing visual observation of the symptoms expressed by farmers. Due to incompletenessin user query or difficulty expressing symptoms correctly, this approach does not accurately recognize the disease. To solve these problems, this paper introduces a new plant disease diagnosis system that supports a multi-modal approach by incorporating imaging data and farmer observation data to improve accuracy in disease detection and reduce ambiguity in user query.

 Keywords: Deep learning, Plant diseases, Ontology, Feature Extraction, Expert Systems.