An Automated Prediction of Plant Diseases Using Leaf Images Based on Efficient Deep Learning and Image Processing

Abstract

In all nations' economies, the agricultural sector can play a vital role through crop production. The discovery of plant diseases could be one of the most significant factors in the preservation of an agriculturally productive country. Crop farmers lose a considerable sum annually since numerous diseases have affected their plants. Various plant parts could have been contaminated by fungi, viruses, and microorganisms, but in this work, the focus is on the discovery and categorization of banana leaf disease. Improved food production quality and reduced economic losses will be achieved with an accurate early prediction of plant leaf disease. Nowadays, deep learning (DL) and artificial intelligence (AI) have been widely applied for the construction of automated systems to discover and categorize banana leaf diseases. In this paper, an automated banana leaf disease prediction system is presented by proposing an enhanced deep learning model. In this work, captured banana leaf images can be pre-processed through combined filters, in which image smoothing and noise reduction are done by a Gaussian filter, followed by a de-noising process that is refined additionally using the Wiener filter scheme. Subsequently, a DL model convolutional neural network (CNN) can be exploited to extract pertinent features. In this paper, an attention and fuzzy logic-based recurrent neural network (AFLRNN) model is proposed and trained with extracted features for efficient prediction and categorization of plant leaf disease. These systems utilize image processing and computer vision techniques to analyze images of plant leaves and identify potential diseases. The proposed DL system is able to efficiently detect leaf disease in plants at an early stage. The presented scheme can learn and extract features from captured plant images to facilitate precise leaf disease discovery and categorization. This kind of early detection of leaf disease is used to provide suitable treatment and reduce crop waste. The experimental results have exposed that the proposed system has attained a above 98 % higher disease prediction accuracy that has been higher to state-of-the-art schemes.