الجامعة التقنية الشمالية

Argos System 2026 is an AI-powered, multi-sensor hardware and software fusion platform engineered to locate human survivors trapped within post-earthquake rubble zones. To overcome the physical limitations of conventional rescue tools-such as thermal cameras blocked by thick concrete structures-the system deploys its multi-sensor array using mechanical, flexible, and articulable robotic arms that physically penetrate deep into narrow debris crevices. The platform captures and cross-validates live biological markers, including non-contact object heat (MLX90614), sub-surface acoustic anomalies, and respiration gas leaks (MQ-135). Operating entirely offline to eliminate cloud latency, a localized Python predictive core evaluates real-time telemetry against a specialized database calibrated with clinical and industrial material limits. By autonomously filtering out structural noise and false environmental heat, Argos provides first responders with immediate, actionable intelligence to optimize heavy excavation and maximize victim survival rates.

Abstract Liver disease is a "silent killer" whose complex early diagnosis often leads to physician burnout. To address this, we developed a comprehensive, AI-Based Liver Disease Diagnosis System that is both highly accurate and user-friendly for clinical settings. Methodologically, we engineered a "multi-model ecosystem." After initially testing the Random Forest algorithm, we selected XGBoost as our primary engine for its superior speed and accuracy, pairing it with strict deterministic logic. Guided by a "Data Quality over Quantity" philosophy, we rigorously cleaned our datasets and removed statistical noise. This approach proved that AI learns far better from small, verified clinical facts than from massive, messy data. As an outcome, this rigorous approach yielded exceptional performance across all stages of the diagnostic pipeline. The system’s "Gate Model" used for initial patient triage achieved an accuracy of 97.41%. For specific disease pathways, the deterministic Fatty Liver (NAFLD) model achieved 100.00% accuracy. The specialized XGBoost engines successfully achieved 95.24% accuracy in predicting Hepatitis C (HCV) complications, 71.43% in assessing HCV mortality status using strictly verified data, 62.50% in classifying complex HCV fibrosis stages, and 94.00% in evaluating Hepatocellular Carcinoma (Liver Cancer) risk. Ultimately, by integrating these models into a seamless full-stack web application, this research provides a highly reliable, practical tool that helps shift liver healthcare from reactive treatment to proactive, life-saving prevention.

Akhdar is an AI-powered smart application designed to combat desertification and improve the survival rate of newly planted trees in arid and semi-arid regions. The system provides personalized recommendations for outdoor trees and indoor plants by analyzing real-time weather, soil, and environmental conditions using Google Gemini 3.1 and a locally curated agricultural dataset. It also includes a computer vision model for plant disease diagnosis from a single image, a 24/7 AI agricultural assistant, weather-based irrigation reminders, and interactive features such as tree naming, gifting, and leaderboards to encourage long-term user engagement. In addition, an integrated marketplace enables users to purchase saplings, gardening tools, and related services. By combining artificial intelligence, localized agricultural knowledge, and community engagement, Akhdar offers an accessible, scalable, and sustainable solution that supports smart afforestation and contributes to the United Nations Sustainable Development Goals (SDGs).

PRISM is an AI system built to understand what is happening to Iraq’s rivers and reservoirs before the damage becomes visible on the ground. It turns satellite observations, hydrological signals, temporal change, and environmental rules into clear water-state intelligence for the Tigris–Euphrates network. PRISM can track abnormal shrinkage, expansion, and risk patterns across monitored segments, then convert them into alerts, recommendations, and investigation-ready evidence. The goal is to give water authorities, researchers, and decision makers a faster way to see where the water system is changing, why the change matters, and what action could be taken. PRISM: www.prismiraq.live

This project addresses the problem of generating realistic, real-time facial animation from a live webcam stream using artificial intelligence, a task at the intersection of deep generative modeling and computer vision. The proposed solution is built around an AI-driven motion-transfer pipeline that combines a PyTorch-based neural network for GPU-accelerated tensor computation with OpenCV for video capture and NumPy for array-based pixel manipulation, drawing on First Order Motion Modeling and related Generative Adversarial Network (GAN) and Latent Diffusion approaches, with the aim of mapping latent motion descriptors extracted from a driving webcam feed onto a target facial identity with minimal latency. The expected outcomes include a working AI-based animation pipeline that re-enacts facial expressions without intrusive physical markers, and an analysis of the main failure modes of current generative motion-transfer models—including latent-space topological distortion, semantic drift and identity leakage, and degraded performance on extreme head poses outside the training distribution—that currently limit the realism of AI-driven animation systems.

With continuos growing process in digital communication, there is increasing in digital world attacks on information systems using social engineering. It is really necessary to perform tasks using artificial intelligent to secure all data and information against phishing threats attackers such as control complicated URLs, incorrect domains, and invalid Secure Sockets Layer SSL certificates. The purpose of this project is using artificial intelligent to design integrated security platform called Cybershield. To implement the platform has been used Python program, Nexus autonomous sentinel artificial intelligent (NASAI), Jarvis artificial intelligent, radar network detection, OpenCV virtual detection, Telegram API, JavaScript libraries (jsencrypt, Crypto-JS). Systems could face risks since browser storage mechanisms, for that Cybershield has improved the efficiency of phishing web resource detection of explainable artificial intelligence and covers the wide range of options to password-based authentication system. This work has built and applied system platform that improved precision to the most important web security.

تُعد هشاشة العظام من الأمراض الشائعة التي تؤثر بشكل كبير في صحة الإنسان وجودة حياته، إذ تؤدي إلى انخفاض كثافة العظام وتدهور بنيتها المجهرية، مما يزيد من خطر الكسور والمضاعفات المرتبطة بها. وتكمن إحدى أبرز التحديات في تشخيص هذا المرض في غياب الأعراض السريرية الواضحة خلال مراحله المبكرة، الأمر الذي يجعل الكشف المبكر والتدخل الوقائي أمراً بالغ الأهمية. وفي هذا السياق، برزت تقنيات التصوير الطبي والتحليل الحاسوبي كأدوات واعدة لدعم التشخيص الانتهازي من خلال الاستفادة من الصور الشعاعية الروتينية المتاحة للمرضى. تركز هذه الدراسة على توظيف أساليب متقدمة لاستخراج الخصائص من صور الأشعة السينية للركبة بهدف تقييم حالة النسيج العظمي الإسفنجي والكشف عن مؤشرات التدهور المرتبطة بهشاشة العظام. وتعتمد الدراسة على دمج طريقتين رياضيتين هما الهندسة الكسيرية (Fractal Geometry) وتحليل نسيج غابور (Gabor Texture Analysis)، لتطوير مؤشر جديد يُعرف باسم "مؤشر اضطراب العظم الإسفنجي" (Trabecular Disturbance Index - TDI). كما يتم استخدام خوارزمية آلة المتجهات الداعمة (Support Vector Machine - SVM) لتصنيف الحالات إلى ثلاث فئات سريرية هي: الطبيعي، وقلة كثافة العظم، وهشاشة العظام. وقد أظهرت النتائج فعالية النهج المقترح، حيث حقق دمج الطريقتين دقة تصنيف بلغت 90.21% على مجموعة بيانات سريرية مكونة من 470 صورة شععية للركبة، مما يؤكد إمكاناته كأداة مساعدة للكشف المبكر عن هشاشة العظام.

Breaking communication barriers between deaf patients and healthcare providers through real-time sign language translation.

Skin diseases are one of the most important health issues of the world but access to specialized dermatological services is limited. In this study, DermaScanAI-20, a custom 20-layer architecture using Convolutional Neural Networks (CNN) is introduced within an intelligent Clinical Decision Support System (CDSS) for automated skin lesion analysis. To overcome the "black box" problem inherent in deep learning, Gradient-weighted Class Activation Mapping (Grad-CAM) was added to create an explainable AI heatmap that points to the specific clinical region that is responsible for the classification. The proposed system is designed to combine automated image preprocessing, deep feature learning, and K-Means clustering for accurate segmentation of the lesions, using a curated dataset of 46,334 dermatological images across fifteen distinct disease categories. The results of the experiments are presented, and it is found that DermaScanAI-20 has an outstanding Overall Testing Accuracy of 92.80%, a Macro-Average Precision of 93.20%, a Macro-Average recall of 92.80% and a harmonic F1-Score of 93.00%. The high validation metrics statistically confirm the strong integrity of the diagnostic process of the system, making it an extremely reliable and transparent screening system for clinical dermatology applications.

ٌThis project addresses a significant problem in human-robot interaction and campus academic orientation. The proposed solution is based on "Dalil," an autonomous, open-source robotic assistant utilizing a hybrid processing architecture that pairs a Raspberry Pi 5 with an ESP32 microcontroller, with the aim of deploying YOLOv8 facial recognition and natural language processing to guide university visitors and independently resolve repetitive academic inquiries. The expected outcomes include achieving highly optimized response latencies under 3 seconds, attaining a speech processing accuracy exceeding 90% in the localized Iraqi dialect via secure API connections, and delivering a substantial reduction in the routine administrative burden imposed on university personnel.

يبرز مشروع "عين النسر" (Eagle Eye) كحلٍ رقميٍ شاملٍ ورائد، يرتكز في جوهره على أحدث تقنيات الرؤية الحاسوبية والذكاء الاصطناعي لإحداث نقلة نوعية في إدارة قطاعين حيويين في العراق، وهما: القطاع المروري: من خلال القراءة اللحظية للوحات المركبات، والرصد الدقيق للمخالفات، وتتبع المركبات المطلوبة أمنياً في الوقت الفعلي. قطاع المواقف الذكية: عبر الأتمتة الكاملة لاحتساب مدد الوقوف والرسوم المستحقة، وإصدار الإيصالات الإلكترونية بسلاسة وشفافية. وتُدار هذه المنظومة المتطورة عبر منصة تشغيلية موحدة، تشكل حلقة وصل تفاعلية تربط بين (الجهات الإدارية، والأجهزة الأمنية والمرورية، والمواطنين) ضمن قاعدة بيانات مركزية رصينة. وقد برهنت هذه المنظومة على كفاءة تشغيلية وسرعة استجابة فائقتين، مما يسهم بشكل فعّال في تقليص الاعتماد على التدخل البشري والارتقاء بمستوى الأداء العام.

This paper presents a design and implementation of Geostring, an end-to-end automated string-art platform that converts any digital photograph into a physical artwork composed of a single continuous thread wound around a circular nail frame. The system comprises three integrated components: an interactive web-based 2D generator employing a greedy chord-selection algorithm in polar coordinates, a live Three.js-based 3D mechanical simulator, and an AI-assisted parameter tuning assistant. The physical drawing machine is built around an Arduino UNO R3 running GRBL firmware, two NEMA 17 stepper motors driven by TMC2225 silent drivers, and a GT2 belt-driven linear arm operating over a 290 mm rotating disc. The greedy algorithm iteratively selects the nail-to-nail chord that maximally reduces the residual image darkness, producing sequences of up to 8,000 chords. Experimental evaluation confirms a disc positioning accuracy of 0.08° mean error, thread-shuttle repeatability of ±0.15 mm, and successful autonomous artwork production. The platform bridges computational art with physical mechatronic fabrication, providing an open, modular, and affordable system applicable to engineering education, creative technology, and intelligent automation research.project addresses a significant problem in (mention the field or domain). The proposed solution is based on (briefly describe the idea or technology) with the aim of (state the main objective). The expected outcomes include (summarize the anticipated results or impact).