Plant Leaf Disease Detection with Feature Extraction and Dense Maxout Forward Harmonic Network

Abstract

Plant diseases are a major problem for farmers and gardeners because they impact the plant’s yield and health. Several Deep Learning (DL) approaches are established by researchers worldwide for detecting plant leaf diseases. However, attaining accurate outcomes at the early stage is challenging. An innovative approach for the detection of disease in plant leaves is proposed in this research. The images are initially obtained from the datasets and the anisotropic diffusion is utilized for denoising in the preprocessing module. Subsequently, the U-Net is exploited for segmenting the plant leaf. The image augmentation is implemented utilizing color change, scaling, and rotation. The feature extraction is processed for mining Local Ternary Pattern (LTP), and Haralick features namely correlation, Inverse Difference Moment (IDM), Angular Second Moment (ASM), and entropy. Finally, the detection of diseases in the plant leaf is performed by the Dense Maxout Forward Harmonic Network (DenMxFH-Net) approach. The DenMxFH-Net is introduced by combining the Harmonic Analysis, Deep Maxout Network (DMN), and DenseNet technique. Furthermore, the DenMxFH-Net approach obtained the lowest False Positive Rate (FPR) of 7.9%. Further, the DenMxFH-Net recorded an accuracy of 92.9%, a True Positive Rate (TPR) of 94.9%, a precision of 91.8%, and an [Formula: see text]-score of 93.3%.

Code and data availability

公開本文の所在を確認できませんでした。非公開または購読が必要な可能性があります。

No evidence-backed public reproduction asset is currently recorded.