EMITTER International Journal of Engineering Technology

An Attention based Vision Transformer for the Detection of Insect Pests in Castor Crop

2026-09-06T06:37:00+00:00

Castor (Ricinus communis L.) is a significant crop valued for its non-edible oil, yet its economic importance is compromised by insect pests causing substantial yield losses of 35-40%. This paper explores the efficiency of utilization of vision transformers for efficient pest classification. We propose CASTIPestViT, a Vision Transformer-based model specifically designed for insect pest detection in castor crops. The model integrates transfer learning and fine-tuning mechanisms, leveraging a pre-trained Vision Transformer (ViT) initially trained on ImageNet1k, and is fine-tuned on a custom dataset of castor insect pests. CASTIPestViT uses the self-attention mechanism of ViTs to capture global and local features of insect pests. The performance of CASTIPestViT is compared with six different pre-trained CNN models. The results obtained by the proposed model achieve a validation accuracy of 97.60% in insect pest detection and outperforming other state-of-the-art models in terms of precision, accuracy, and f1-score. The model offers a robust solution in early-stage insect pest detection to reduce yield losses. The efficiency and accuracy of the model make it suitable in sustainable crop management and smart agriculture systems for yield optimization.

Accurate Automatic Recognition of Iraqi License Plates Using YOLOv11n and Enhanced OCR Techniques

2026-09-06T06:36:42+00:00

Automatic License Plate Recognition (ALPR) systems are widely utilized for traffic monitoring, law enforcement, and security applications. While previous ALPR techniques have demonstrated high accuracy on uniform license plate formats, recognizing license plates in many countries particularly in Arabic-speaking regions remains a significant challenge due to the complex structure and diverse designs of the plates. Moreover, visual similarities between certain English letters and Arabic numerals often lead to misinterpretations by Optical Character Recognition (OCR) systems. This study proposes an efficient ALPR framework tailored for the new Iraqi license plates, employing image processing and deep learning (DL) techniques. The system integrates the YOLOv11n object detection model for accurate license plate localization, and OCR for character recognition. To address the OCR misclassification issue, particularly those caused by Arabic-English character similarities, the OCR component is enhanced using Character Index Checking (CCI) and Regular Expression Patterns (REP), resulting in more stable and accurate recognition outputs. The proposed method was evaluated on a real dataset of Iraqi license plates under varying illumination, viewpoints, and background complexity. Experimental results demonstrate strong end-to-end ALPR performance, achieving 100% plate detection accuracy, 99.8% recognition recall, and 99.5% mAP@0.5, confirming the robustness and practical effectiveness of the proposed system.

Exploring YOLO-Based Deep Learning Approaches for Fish Detection in Intelligent Aquatic Monitoring Systems

2026-09-06T06:36:23+00:00

The Advancements in precision aquaculture demand robust visual monitoring systems capable of accurate, real-time fish detection in complex underwater environments characterized by turbidity, occlusion, and dynamic illumination. While YOLO (You Only Look Once) architectures have demonstrated high efficiency in object detection tasks, their comparative performance for underwater fish detection remains underexplored, particularly across recent variants such as YOLOv5, YOLOv8, and YOLOv11. This study presents a systematic evaluation of three state-of-the-art YOLO models using a curated GlowFish dataset consisting of 533 annotated images across three fluorescent species. Data were acquired under controlled but visually diverse conditions using multi-angle imaging and standardized illumination. A uniform training pipeline, consistent annotation using the COCO format, and identical hyperparameters were applied across models to ensure fair benchmarking. Key evaluation metrics include precision, recall, mAP@0.5, and mAP@0.5:0.95. Experimental results reveal that YOLOv5 achieved the highest precision (0.963) and mAP@0.5 (0.967), while YOLOv8 delivered superior recall (0.930) and more balanced detection across species classes. YOLOv11 demonstrated architectural potential but showed greater sensitivity to class imbalance and reduced confidence stability. Visual analysis and confusion matrices further confirmed model-specific trade-offs in classification reliability and localization precision. This work contributes critical empirical insights into the selection of YOLO architectures for intelligent aquaculture systems, offering practical guidance for real-time aquatic monitoring deployments. Future research will extend this framework to multi-species, multi-environment datasets, integrate spatiotemporal behavioral tracking, and investigate deployment on resource-constrained edge-AI platforms, advancing the field toward interpretable and autonomous aquatic monitoring solutions.

Supporting Independent Prayer in Alzheimer’s Patients with an EEG-Enabled BCI System

2026-09-06T06:36:05+00:00

Alzheimer's disease (AD) significantly impairs cognitive functions, making independent activities, including Muslim prayers, challenging for patients. This study introduces an innovative Brain-Computer Interface (BCI) system leveraging Electroencephalography (EEG) signals to facilitate prayer practices for individuals with AD. Utilizing a 5-channel EEG headset to monitor attention levels, our system detects alpha and beta wave patterns to assess user focus. When attention diminishes, the system provides guidance to the next prayer step, ensuring continuity and support. Additionally, motion detection technology captures physical movements associated with prayer, enabling the classifier to learn and recognize different prayer postures from a dataset of two individuals. This approach aids AD patients in maintaining religious practices independently while significantly enhancing their psychological well-being by fostering autonomy and spiritual fulfillment. Our findings suggest that integrating EEG-based BCI systems with motion detection offers a promising avenue for supporting daily activities and improving quality of life for individuals with cognitive impairments.

Area-Efficient Concryption for Secure Data Transmission

2026-09-06T06:35:12+00:00

As computer networks become more connected, it is more important than ever to protect users' data and privacy. Especially because modern cyberattacks are so advanced, cryptography is a key way to keep user data private, real, and safe. However, since cryptographic methods do not reduce file sizes, concryption techniques are employed to optimize storage and enhance bandwidth for secure and efficient data transmission. The concryption method is implemented using two of the most flexible and efficient symmetric block ciphers: The concryption method uses the Blowfish and Twofish symmetric block ciphers, along with compression techniques such as LZW and MTF. We propose an area-efficient encryption algorithm that is better than the Twofish algorithm. The proposed architecture is a suitable fit for IoT devices, embedded systems, and hardware platforms that don't need much power (0.135 mW) and don't need much space (1,197 gates). Therefore, the proposed encryption architecture is a good balance between the cost of hardware, the effectiveness, and the security of lightweight cryptographic applications. The proposed encryption technique is applied to the concryption process, where compression before encryption results in a better compression ratio compared to encrypting the data first, thereby reducing bandwidth usage and enhancing transmission speed for secure data transfers.

The Ontology Driven E-Commerce Chatbot for Fashion Industry

2026-09-08T06:38:08+00:00

Conventional chatbots in fashion e-commerce rely on keyword matching and lack the semantic structure needed to correctly handle multi-entity product queries. This leads to inaccurate, context-blind responses that frustrate users and reduce platform engagement. This paper proposes and evaluates an ontology-driven chatbot system for fashion e-commerce and compares it directly against a non-ontology baseline built on the same dataset. Both systems were developed using Google Dialogflow for natural language processing, with product data collected through Python web scraping from the Alkaram fashion website. The ontology knowledge base was engineered in OWL 2 using Protégé and queried via SPARQL, with a class hierarchy covering product categories, attributes, and brand entities. To evaluate both models, five analysts conducted structured 10-minute conversations using a fixed set of 50 product queries, scoring each response across seven criteria, including accuracy, context retention, and grammatical quality, on a 0 to 9 scale. The ontology-based system achieved 88% accuracy compared to 67% for the non-ontology model, a 21% improvement. It also outperformed across all other criteria, with inter-analyst agreement within 0.5 points confirming evaluation reliability. These results show that integrating an OWL/SPARQL ontology into a Dialogflow-based chatbot meaningfully improves product query handling in fashion e-commerce, and the approach is practical enough to scale to other retail domains.

An iterative LQR control method for 6-DOF UAV quadcopter under disturbance and payload variation

2026-09-06T06:35:30+00:00

This paper proposes an optimal control method based on the iterative LQR aimed at enhancing the trajectory tracking accuracy of a 6 DOF quadcopter UAV under external disturbances and varying payloads. The main distinguishing feature of the proposed method is the iterative optimization mechanism of the weighting matrices Q and R through the logarithmic normalization of error metrics, including ITAE, ISE, and RMSE, which automatically adjusts the controller based on the system's actual performance. The effectiveness of the method is evaluated through simulations in MATLAB across various flight trajectories and scenarios, including ideal conditions, disturbances, and combined disturbances with varying payloads. The simulation results show that the iterative LQR method significantly improves control quality compared to traditional PID and LQR controllers, with error reductions ranging from 6.6% to 41.8%, particularly evident in the x and y position axes and the yaw angle ψ. Furthermore, preliminary experimental results on quadcopter UAV hardware show that the proposed iterative LQR algorithm helps improve control quality and enhance flight attitude stability under low wind disturbance and varying payloads. These results demonstrate the adaptability and robustness of the proposed method under the complex operating conditions of UAVs.

Modeling and Forecasting Piezoelectric Energy Harvesting Using Deep LSTM–ANN Architectures

2026-09-06T06:35:47+00:00

Piezoelectric energy harvesting (PEH) enables maintenance-free micro-power generation for autonomous sensing and ultra-low-power electronics by converting ambient mechanical excitation into electrical energy. Despite substantial progress in piezoelectric materials and device structures, forecasting PEH electrical outputs remains difficult because the response is nonlinear, excitation is stochastic, and performance can drift under repeated loading. This paper proposes a hybrid deep learning architecture that integrates Long Short-Term Memory (LSTM) and Artificial Neural Network (ANN) components to forecast voltage, current, and power from footstep-driven PEH time-series data. The dataset is constructed by sampling the harvester voltage under controlled walking-induced excitation and organizing the continuous signal into supervised samples using a sliding-window scheme; features are normalized and paired with future targets for multi-output regression. The model is trained and evaluated against standalone LSTM, standalone ANN, and classical forecasting baselines using RMSE, MAE, MSE, and R². Experimental results show high voltage prediction accuracy (R²=0.9896, RMSE = 0.0035, MAE = 0.0022), while current and power are predicted with acceptable performance consistent with their higher noise sensitivity and nonlinear coupling. These findings indicate that combining temporal memory with nonlinear regression improves forecasting stability for PEH outputs within the defined experimental setting and provides a practical basis for energy-aware scheduling and monitoring in self-powered sensing applications. Future work will extend the dataset to broader excitation conditions and incorporate uncertainty-aware modeling for robust edge deployment.

Particle Swarm Optimization–Tuned Random Forest for Multi-Class Mental Health Sentiment Classification from Textual Data

2026-09-08T06:37:02+00:00

Mental health sentiment classification from textual data has attracted increasing attention as a computational approach to support large-scale psychological assessment; however, multi-class classification remains challenging due to noisy text, class imbalance, and semantic overlap among categories. This study proposes and evaluates a machine learning framework for seven-class mental health sentiment classification that integrates enhanced text preprocessing with lemmatization, data augmentation via back-translation, TF-IDF feature extraction, and systematic model evaluation across multiple classifiers, including Logistic Regression, Decision Tree, Random Forest, Gradient Boosting, K-Nearest Neighbors, AdaBoost, and XGBoost, under three hyperparameter tuning strategies: Grid Search, Random Search, and Particle Swarm Optimization (PSO). Experimental results indicate that ensemble-based models consistently outperform single classifiers, with the PSO-optimized Random Forest achieving the best numerical performance, attaining an accuracy of 0.933, a macro F1-score of 0.923, and a ROC AUC of 0.989, demonstrating strong generalization and balanced class-level performance despite dataset imbalance. These findings confirm that the combination of robust preprocessing and metaheuristic-based hyperparameter optimization significantly enhances multi-class mental health sentiment classification and supports its potential use as a scalable decision-support tool for large-scale mental health screening, while not intended to replace clinical diagnosis.