Journal of Tomography System and Sensor Application

Grape Leaf Disease Detection Using Convolutional Neural Network

2025-06-13T06:24:29+00:00

This project focuses on developing a plant disease detection system using Convolutional Neural Networks (CNN) to address the critical challenge of identifying plant diseases early in agriculture. The proposed system leverages image analysis to classify diseases such as Grape Black Rot, Leaf Blight, and healthy conditions in grape leaves. Utilizing AlexNet architecture in MATLAB, the model processes a dataset of 500 leaf images (70% for training, 30% for testing) with image preprocessing techniques like resizing and normalization. The methodology involves designing a MATLAB-based GUI for user interaction, allowing image uploads, disease detection, affected area analysis, and remedy suggestions. Model performance was evaluated on multiple metrics, achieving an overall accuracy of 99.3% on the validation dataset. Tests on 60 samples consistently demonstrated high prediction confidence (96.42%-100%) and accurate classification of healthy and diseased leaves. Quantitative analysis of the affected area using clustering revealed detailed insights into disease severity, supporting effective decision-making. This system shows strong potential for real-time agricultural applications, contributing to sustainable farming practices and enhancing food security. Future enhancements include integrating mobile platforms for broader accessibility.

Potential of Portable Solar Powered Electrolysis System for Off-Grid Hydrogen Generation

2025-07-18T01:00:17+00:00

The growing interest in hydrogen as a clean fuel alternative stem from the adoption of sustainable and decentralized energy systems. Portable solar-powered hydrolysis systems show promise for off-grid hydrogen generation because they can serve emergency relief and rural electrification needs and mobile scientific missions. The research investigates theoretical capabilities of these systems through both literature-based analysis and performance modeling. Photovoltaic energy generation is part of the suggested system framework, which is followed by proton exchange membrane (PEM) electrolyzers for compact hydrogen storage and water electrolysis. The performance calculations use solar irradiance data together with electrolysis energy requirements. The research indicates that portable solar-hydrolysis systems have the potential to operate in low-power remote applications with zero emissions despite their current scalability and hydrogen storage challenges

A Lightweight Surveillance System for Human Re-Identification Using Traditional Image Processing Techniques

2025-06-13T06:12:11+00:00

Surveillance systems play a pivotal role in enhancing security, especially in large-scale environments such as academic campuses. While modern re-identification systems typically employ deep learning techniques, these models demand significant computational resources and extensive labeled datasets, limiting their applicability in real-world low-resource environments. This study presents the Integrated Campus-Activities Monitoring System (ICAMS), a lightweight, real-time surveillance system for human re-identification based on traditional image processing methods. The proposed system integrates a revised Structural Similarity Index (SSIM) and a novel Luv Similarity metric within the CIELUV color space, eliminating the need for computationally expensive deep learning models. Experimental evaluations show that the system achieves satisfactory accuracy and robustness in appearance-based human tracking while maintaining low hardware requirements. These results demonstrate the potential for deploying cost-effective surveillance systems capable of real-time performance in environments with limited infrastructure.

Non-Destructive Prediction of Papaya Maturity Level Using Near-Infrared Spectroscopy and Artificial Neural Network

2025-06-26T07:57:03+00:00

This paper presents a non-invasive technique for evaluating the maturity of Carica Papaya L. by predicting its Soluble Solid Content (SSC) using Near-Infrared (NIR) spectroscopy integrated with an Artificial Neural Network (ANN) model. Traditional destructive methods using refractometers hinder quality preservation and consumer usability. In contrast, the proposed approach utilizes NIR spectral reflectance data, pre-processed with Savitzky-Golay (SG) smoothing and derivatives, and is analyzed via a nonlinear ANN regression model. Experimental results based on 49 papaya samples show high predictive accuracy (R² is 0.9063 for training, 0.8768 for testing; RMSE is 0.4406 and 0.7047 respectively) using second derivative data. This study demonstrates the feasibility of portable NIR systems for real-time fruit maturity classification, supporting broader applications in agricultural and supply chain contexts.