Plant Leaf Disease Detection with Multivariable Feature Selection Using Deep Learning AEN and Mask R-CNN in PLANT-DOC Data

Nithyanandh Selvam^1* and Eldho Konnammanayil Joy²

¹Department of MCA, PSG College of Arts and Science, Coimbatore, Tamilnadu, India.

²Department of Computer Science, Mary Matha Govt Aided Arts and Science College, Mananthavady, Kerala, India.

Corresponding Author E-mail:nknithu@gmail.com

ABSTRACT: Objectives: To suggest a new AI-based model to detect plant leaf diseases at an early stage in order to maximize crop yield. Deep learning with multi-variable feature selection method is employed to boost the accuracy rate of detection and classification. Methods: The artificial intelligence-based mask RCNN is utilized to extract all the multivariable features in plant leaves to predict the type of disease at an early stage. DAEN is employed to denoise the image and learn the leaf image representations from unlabeled data to improve the classification accuracy. Combinational methods like local binary patterns, color histograms, and shape descriptors are employed to identify local and global features of the plant leaf. The PLANT-DOC dataset is used for this research study, which includes 2,590 leaf images and 17 classes of disease with the target attributes of healthy and diseased leaves. The LeafNET architecture is used for pre-processing and to analyze the significant spots in leaf images. To evaluate the performance of the proposed AI-based Mask-RCNN, a MATLAB tool is used, and the results are compared with the prevailing models such as ACO-CNN, I-SVM, KNN, and DL-RPN. Findings: The suggested AI Mask R-CNN model outperforms the existing methods with proven results of 95.06% accuracy, 96.7% sensitivity, 97.24% specificity, 96.4% TPR, 95.91% TNR, 9% FPR, 8.4% FNR, and 94.57% F1 Score, 94.96% detection speed which clearly shows that the model has the ability to classify and detect leaf disease in a robust manner. Novelty: The outcome of the AI Mask-RCNN model enhances the accuracy rate in terms of plant leaf disease detection and classification at an early stage, which helps the agricultural sector maximize crop yields. In terms of performance evaluation, the proposed model outperforms the shortcomings of the existing leaf disease detection models such as ACO-CNN, I-SVM, KNN, and DL-RPN.

KEYWORDS: Auto Encoder Networks; Artificial Intelligence; Classification; Data Processing; Image Segmentation; Mask R-CNN; Plant Leaf Disease Detection

[ HTML Full Text]