Abstract
India has significant food production potential that can be used to meet the needs of our planet. The north region of India, in particular, is suitable for citrus fruit, which is of great interest to regional businesses. To fully utilize these capabilities, early detection of diseases, pests, and nutritional deficiencies, as well as timely management of these issues, must be ensured. This necessitates the development of solutions that aid in the early detection of conditions. Because of the current relevance of deep learning for image analysis, it was decided to propose a system for field diagnosis that combines this concept with digital image processing. Thus, in this work, a system is proposed that, based on visible light spectrum images of its leaves, allows the detection in the field of conditions in citrus crops, providing a rapid response without the high costs and complications of other methods. The system is designed with deep learning and image processing in mind: a convolutional neural network is trained using transfer learning and data augmentation, among other techniques, and then used to create an android application that can take images and diagnose them. For diagnosis, the affections are classified into seven categories. The main findings include accuracy rates of more than 90% when using the classification model in conjunction with the rejection threshold technique on the test dataset, as well as the model’s successful deployment with barely noticeable execution times.
Subject Classification: