An efficient adaptive feature selection with deep learning model-based paddy plant leaf disease classification

• Published in: Multimedia tools and applications Vol. 83; no. 8; pp. 22639 - 22661
• Main Authors: Dubey, Ratnesh Kumar, Choubey, Dilip Kumar
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.03.2024; Springer Nature B.V
• Subjects: Adaptive algorithms; Blight; Color; Computer Communication Networks; Computer Science; Crops; Data Structures and Information Theory; Deep learning; Economic impact; Feature extraction; Feature selection; Image classification; Machine learning; Medical imaging; Multimedia Information Systems; Optimization; Optimization algorithms; Plant diseases; Signs and symptoms; Special Purpose and Application-Based Systems; Support vector machines; Deep learning; Paddy leaf plant; SV-RFE; Rain optimization algorithm; Median filter; ABi-LSTM
• DOI: 10.1007/s11042-023-16247-3
• ISSN: 1573-7721; 1380-7501

Agriculture is the essential source of national income for some nations including India. Infections in crops/plants are serious causes of reduced quantity and quality of production, resulting in economic loss. Therefore, the detection of diseases in crops is very essential. Plant disease symptoms are evident in different parts of plants. However, plant leaves are commonly used to diagnose infection. Therefore, in this paper, we focus on automatic leaf disease detection using the deep learning model. The presentedmodel consists of four phases namely, pre-processing, feature extraction, feature selection, and classification. At first, the captured paddy leaf images are converted into an RGB color modelthe median filter is used to remove the noise present in the green band. Then, the texture and color features are extracted from the green band. After the feature extraction, important features are selected using a combination of machine learning and optimization algorithm. Here, initially, the features are selected using support vector machine-recursive feature elimination (SV-RFE) and an adaptive rain optimization algorithm (ARO). Then, the common features are selected. The selected features are given to the adaptive bi-long short-term memory (ABi-LSTM) classifier to classify an image as Blast disease, Bacterial Leaf Blight disease, Tungro, or normal image. The efficiency of the presented technique is estimatedbased on the accuracy, sensitivity, specificity, and performance compared with state-of-the-art works.