Citrus Leaves Disease Detection and Classification on Image Recognition Using Deep Convolutional Neural Networks: One Health Perspective

Abstract

Agriculture production is a crucial economic backbone for any country and is vital in meeting human food needs. At the same time, plant disease poses a significant threat to this sector, leading to decreased yields and heavy losses. Automated systems for disease detection and classification can aid in combating this issue and promoting growth and development. In recent years, deep learning approaches have demonstrated promising results in various artificial intelligence tasks, specifically in the Smart Agriculture domain. Smart Agriculture applications include Plant disease detection, water and soil management, crop distribution, crop cultivation, fruit counting, and yield prediction. This paper presents an integrated and enhanced approach for detecting citrus leaf disease detection using a deep convolutional neural network. The proposed model can distinguish healthy citrus leaves from seven common diseases: bacterial spot, black spot, canker, citrus powdery mildew, greening, melanose, and health. The proposed model extracts the complementary features by incorporating multiple hidden layers and using data augmentation for improved image recognition and classification. The proposed model is tested against other deep learning models on the citrus and Plant Village dataset and outperformed previous studies in various performance measurement metrics. With a test accuracy of 97.66%, our model serves as a reliable tool for citrus plant disease detection.