An Ensemble Approach for Tulsi Plants Diseases Detection using MLP Mixer and LSTM

DOI: https://doie.org/10.0210/Jbse.2025142029

PRABHA.B , Dr. K.KAVITHA

Keywords:

Anisotropic Diffusion Filtering, ensemble, Convolutional Neural Network, Support Vector Machine.

Abstract:

Tulsi belongs to the family Lamia ceae and is considered as a holy plant with multipurpose uses in agri horticulture and medicinal field. Diseases in tulsi plants usually develop unknown to the farmer, thus timely and proper diagnosis plays an important role in obtaining high yield and quality plants. This research outlines the use of advanced preprocessing, training, and classification methodologies for detecting tulsi plant diseases using an ensemble-based model. The preprocessing phase of the Feature-Enhancing technique uses the Anisotropic Diffusion Filtering (ADF) for improving contrast and edge sharpness of images and getting rid of excessive image noise. After that, the Convolutional Neural Networks (CNN) is employed for feature extraction since the network is capable of learning hierarchical features from plant images. The extracted features are then fed to LSTM networks as well as MLPs to capture both temporal relatedness and other more realistic nonlinear temporal characteristics. In classification task, a SVM is used to accurately classify the diseased and healthy leaves. The proposed ensemble method outperformed the individual models in all the experiments, offering high accuracy for disease diagnosis. This research presents a more reliable and effective approach for real-time diagnosis of tulsi plant diseases. 

PDF