Image recognition of plant diseases based on Siamese Networks

Abstract

Plant disease is the main harm to the normal growth of crops. The early identification and diagnosis of plant disease is the core of plant disease management. The traditional disealse identification method has the disadvantages of low efficiency and strong subjectivity through human eye observation or infectivity experiment. However, artificial intelligence related technologies are widely used in agriculture, especially in plant disease identification and fruit quality detection. In this paper, digipathos, the public data of plant pathology provided by EMBRAPA, is used as the research object, and a method for identifying different plant diseases based on Siamese Networks and image processing is proposed. The experiment uses 9 crop categories and 50 corresponding disease categories, a total of 33567 pictures as the training set, and the experimental evaluation uses a total of 3356 pictures as the verification set, with a recognition accuracy of 80.6%, The effectiveness of this method in plant disease identification is verified, which is of great significance to improve agricultural products and ensure the green and sustainable development of agricultural ecological environment. (Abstract)