The hybrid feature incorporated dual deep learning architecture for the automatic jasmine plant disease detection and classification

Abstract

Agriculture is the strength of our nation, and its pro-duction rate is also important for feeding food to all. The crop production rate is more significantly affected by various diseases. Currently, these diseases are iden-tified and classified using a visual inspection method, which is not suitable for larger crop fields. Therefore, a machine-based, systematic approach is needed to identify various diseases that occur on the leaves of the jasmine plant. The present article develops and proposes a computer-based systematic approach to the detection and diagnosis of jasmine plant diseases using the dual deep learning method. The proposed classifier consists of a general adversarial network (GAN) module and a proposed convolutional neural network (CNN) archi-tecture for diagnosing diseases in the jasmine plant. The GAN module extracts the features from the data-augmented jasmine plant leaf image, and the CNN module performs the disease classification process. The proposed CNN module contains both lower and higher-order kernels that produce the hybrid features, which are further classified by the CNN architecture. The proposed classification approach is validated on the set of jasmine plant leaf images. The simulation is carried out using MATLAB software, and the results of the plant leaf classification system are compared with state-of-the-art models in terms of crop sensitivity, crop specificity and accuracy.