Optimal Feature Selection and Classification Using Convolutional Neural Network-Based Plant Disease Prediction

Abstract

Plant disease diagnosis remains a maj or unsolved problem in the changing scientific world. One of the most important agricultural sectors is growing plants with artificial intelligence support. Image analysis and classification techniques are aimed at predicting plant diseases. So far, well-versed farmers or plant specialists can only diagnose plant diseases. Similar results have been shown for almost different plant diseases in the methods thus introduced. To solve such a problem, we have developed an improved Convolutional Neural Network (CNN) method to identify plant diseases faster and better. Initially, the preprocessing task was done by using a wrapping filter. After that, the best features of the plant disease can be selected from the Logistic Decision Regression (LDR). LDR feature selection is used to reduce the classification problem to selecting features of medicinal plants. Leaves are commonly used to identify medicinal plants, branches, flowers, petals, seeds and whole plants for use in automated procedures. Automated disease detection methods are developed based on changes in plant foliar disease states. Convolutional neural networks (CNNs) are the most accurate of the complex feed- forward neural networks in image classification and recognition. By estimating the results, various images can be training have efficient image recognition functions with accuracy and strong reliability.