Feature extraction and classification based crop disease detection using deep neural network and ensemble convolutional neural network architecture

Abstract

Crop disease identification is crucial for avoiding production losses and lowering the amount of agricultural items produced. Machine learning-based approaches can be utilized to solve these types of problems. Deep Learning Techniques, which are commonly employed in image processing, have recently been involved in a number of agricultural applications. This research suggests detecting agricultural diseases using feature extraction and classifying those using deep learning algorithms. For the extraction of the segmented feature, we employ a pre-training model-based Deep Neural Network. The segmented feature will then be categorized using an ensemble classifier composed of Visual Geometry Group (19) and Inception V3++ architecture. This Ensemble Convolution Neural Networks (EN-CNN) classification will increase illness detection accuracy while minimizing time. Pre-trained Deep Convolution Neural Network (PDCNN) and EN-CNN are used to classify damaged and healthy leaves in a dataset of healthy and diseased leaves. The suggested model's resilience is empirically shown for crop disease detection in tomatoes and grapes. The suggested approach evaluated parameters with a 97 percent accuracy rate, a 98 percent precision rate, a 98 percent recall rate, and a 98 percent F-1 score rate.