Skip to main content

2024 | Book

Innovations in Smart Cities Applications Volume 7

The Proceedings of the 8th International Conference on Smart City Applications, Volume 2

Editors: Mohamed Ben Ahmed, Anouar Abdelhakim Boudhir, Rani El Meouche, İsmail Rakıp Karaș

Publisher: Springer Nature Switzerland

Book Series : Lecture Notes in Networks and Systems


About this book

Many cities in the developed world are undergoing a digital revolution, and have placed the "smart city" on their list of priorities. Smart cities use technological solutions such as Internet of Things, AI, 5G, Big Data, Cloud computing, Smart Grid, as well as all the emerging technologies of the digital era, to improve the management and efficiency of the urban environment. The aim is to make residents happier, healthier, smarter and more prosperous, and to make the city greener, cleaner, more sustainable, more responsible, more functional, more resilient, and more competitive.

Enhanced by extensive research studies and carried out under the guidance of international scientific experts in the field. This book explores various papers related to smart cities, including digital twins, geo-smart information systems, education, healthcare, economy and digital business, building and home automation, environment and agriculture, and information technologies and computer science.

Table of Contents


Smart Agriculture

Plant Disease Classification and Segmentation Using a Hybrid Computer-Aided Model Using GAN and Transfer Learning

Plants are essential for life on earth, providing various resources and are helpful in maintaining ecosystem balance. Plant diseases result in reduced crop productivity and yield. Manual detection and classification of plants diseases is a crucial task. This research presents a hybrid computer aided model for plant disease classification and segmentation. In this research work we have utilized PlantVillage dataset with 8 classes of plant diseases. The dataset was annotated using a Generative Adversarial Network (GAN), four transfer learning models were used for classification, and a hybrid model is proposed based on the pretrained deep learning models. Instance and semantic segmentation were used for localizing disease areas in plants, using a hybrid algorithm. The use of GAN and transfer learning models, as well as the hybrid approach for classification and segmentation, resulted in a robust and accurate model for plant disease detection and management in agriculture. This research could also serve as a model for other image classification and segmentation tasks in different domains. Proposed hybrid model achieved the promising accuracy of 98.78% as compared to the state-of-the-art techniques.

Khaoula Taji, Yassine Taleb Ahmad, Fadoua Ghanimi
Water Amount Prediction for Smart Irrigation Based on Machine Learning Techniques

Water is a critical resource that needs to be perfectly managed in the agriculture field to achieve high crop production with minimum water usage and without wastage. In this paper, we proposed a smart irrigation solution using different Machine Learning (ML) models to predict the daily irrigation water amount for the cucumber crop. The various used ML takes the plant’s environmental conditions parameters as input to predict the suitable amount of water as output. The results showed that the Support Vector Regression was the best model that gave the highest coefficient of determination ( $$R^2$$ R 2 score $$\,\approx \,60$$ ≈ 60 %) with the smallest Mean Squared Error value (0.28).

Hamed Laouz, Soheyb Ayad, Labib Sadek Terrissa, M’hamed Mancer
Smart Irrigation System Using Low Energy

The Internet of Things (IOT) makes all areas of our daily lives more comfortable. By connecting physical objects to the internet, without human intervention. The development of new intelligent systems in the field of agriculture has strengthened agricultural production, made it an altarpiece and reduced the cost of production. The purpose of this article is to realize a new fully autonomous model capable of functioning correctly in the agricultural field, especially in places where there is no internet connection, and electricity. Our new model uses a solar panel, an ESP32 microcontroller, and LORA protocol to irrigate agricultural fields to ensure good water management. The performance of our new model will be measured in terms of energy savings. This new model will improve new techniques using IOT in agriculture.

Kamal Elhattab, Karim Abouelmehdi, Abdelmajid Elmoutaouakkil, Said Elatar

Smart Models

Advancing Crop Recommendation Systems Through Ensemble Learning Techniques

In order to assist farmers in selecting the most suitable crops based on environmental characteristics, this article introduces a novel system for crop recommendation that leverages machine learning techniques, specifically ensemble learning with a voting classifier. A comprehensive analysis of prior research in the field of crop recommendation systems reveals the limitations and challenges of previous approaches, particularly their low accuracy. To address these shortcomings, the proposed system incorporates a voting classifier that amalgamates the performance of various machine learning models, while taking into account the perspectives of all participating models. By harnessing the collective intelligence of these models, this approach aims to mitigate the limitations of previous methods and provide more dependable and precise crop recommendations. The results demonstrate the system’s capacity to generate highly accurate recommendations, with the ensemble learning approach achieving an accuracy rate of 99.31%. This empowers farmers to optimize their agricultural practices and maximize crop yields, enabling them to make informed decisions for sustainable and efficient farming.

M’hamed Mancer, Labib Sadek Terrissa, Soheyb Ayad, Hamed Laouz, Noureddine Zerhouni
Technology to Build Architecture: Application of Adaptive Facade on a New Multifunctional Arena

Adaptive façades (AFs) can adapt to changing boundary conditions according to short-term weather fluctuations, diurnal cycles, or seasonal models. The behaviour of indoor environment and the global comfort of a building are strictly dependent on the façade: traditional façades behave statically towards external and internal climate conditions. The objective of this study is to design an adaptive facade system with different layer functions, ensuring the thermal and visual comfort of the various indoor environments and controlling the incident solar radiation.Furthermore, by incorporating second and third generation photovoltaic cells into the adaptive envelope, it is possible to store and produce renewable energy to integrate the “invisible” photovoltaic technology in the building (BIPV). This facade configuration fulfils the performance requirements of the case study presented in this paper: a new multi-functional arena in Paris. This building has a high number of users and different areas of use inside it, so greater flexibility is also required by the envelope. Moreover, the architectural characterisation of the adaptive envelope contributes to establish the building as a new landmark for the neighbourhood and the city.The paper proposes a methodological process that has led to the technological and architectural definition of the envelope element using parametric modelling. By developing a model on Rhinoceros and Grasshopper, it is possible to control configuration and mechanism of the façade, depending on incident solar radiation and the changes in surface temperature.

Alessandra Annibale, Emily Chiesa, Giulia Prelli, Gabriele Masera, Andrea Kindinis, Arnaud Lapertot, Davide Allegri, Giulio Zani
Effectiveness of Different Machine Learning Algorithms in Road Extraction from UAV-Based Point Cloud

This study presents the evaluation of seven different machine learning (ML) models to classify road surface from point cloud. The study begins with converting two-dimensional images collected from unmanned aerial vehicles (UAV) flights to three-dimensional (3D) point cloud. Seven different ML models, namely, Generalized Linear Model, Linear Discriminant Analysis, Robust Linear Discriminant Analysis, Random Forest, Support Vector Machine with Linear Kemel, Linear eXtreme Gradient Bossting, and eXtreme Gradient Boosting, were developed under different training samples. Finally, road surface were classified from 3D point cloud using developed ML models. To assess the performance of the ML models, manually extracted road surfaces were compared with the ones obtained from ML models. Generalized Linear Model produces the most accurate classification results in a shorter processing time. On the other hand, Linear eXtreme Gradient Boosting and eXtreme Gradient Boosting models produce less accurate road classification in a longer processing time. The classification accuracies of other ML models are between these.

Serkan Biçici
A Comparative Analysis of Memory-Based and Model-Based Collaborative Filtering on Recommender System Implementation

Today, several successful companies like Uber, Airbnb, and others have adopted sharing economy business models. The increasing growth of websites and applications adopting this model pushes companies to develop differentiation strategies. One of the strategies is to use emerging technologies to offer a better customer experience. Recommender systems (RSs) are AI-based solutions that can provide customized recommendations. To implement an RS in a sharing economy platform, this study intends to compare the performance of two recommendation-system approaches based on their accuracy, computation time, and scalability. The Netflix dataset was used to compare matrix factorization and memory-based techniques based on their performances using offline testing. The results of the study indicate that memory-based methods are more accurate for small datasets but have computation time limitations for large datasets. Single-value decomposition methods scale better than memory-based algorithms.

Karim Seridi, Abdessamad El Rharras
Critical Overview of Model Driven Engineering

Model-driven engineering (MDE) is gaining favor as a method for creating complex software systems that is both effective and efficient. MDE places a strong focus on using models to represent the various facets of a software system. These models serve as the foundation for creating executable code. Even though MDE has proved effective in some situations, there are still difficulties with the method, such as the difficulty of modeling specific system components and the expense of maintaining the models as a project grows. In this article, we provide a critical analysis of MDE and discuss how it may develop in the future in terms of several concepts. We first consider the drawbacks of conventional MDE methods before looking at alternative remedies that could improve model precision and automate some components of the paradigm. The analysis that was done briefly demonstrates the possible advantages of incorporating AI techniques in order to enhance the MDE process.

Yahya El Gaoual, Mohamed Hanine
A Synthesis on Machine Learning for Credit Scoring: A Technical Guide

Machine learning is a broad field that encompasses a wide range of techniques and algorithms that can be used to perform a wide variety of tasks. The selection of an appropriate algorithm to be used in a particular application can be challenging due to the complexity of the various techniques that are available as well as the high cost of implementing and debugging sophisticated models. In this paper, we examine the use of multiple machine learning algorithms on an Australian dataset that consists of a collection of loan applications from prospective borrowers with differing credit scores. Our goal is to provide comprehensive information about the performance of these models in order to assist financial firms in selecting the most effective model for their needs. To accomplish this goal, we compare the performance of the various models on the classification task and identify the most accurate and effective model based on the overall obtained performance. Our results suggest that XGBoost Classifier, Bagging Classifier, and Support Vector Machine are among the most effective models that can be used for this task based on their superior accuracy when compared to other machine learning algorithms.

Siham Akil, Sara Sekkate, Abdellah Adib
Enhancing Writer Identification with Local Gradient Histogram Analysis

Writer identification is a critical aspect of document analysis and has significant implications in various domains, including forensics, authentication, and historical research. In this article, we propose a novel approach for writer identification using gradient angle histograms collected from neighboring pixels. By calculating the histogram of gradient angles from different locations of neighboring pixels, we effectively capture the writer’s unique style and nuances. Our experimental study demonstrates promising results on the two datasets BFL and CERUG, showcasing the potential of our proposed technique in improving the state-of-the-art methods in writer identification.

Abdelillah Semma, Said Lazrak, Yaâcoub Hannad
Solving a Generalized Network Design Problem Using Hybrid Metaheuristics

Metaheuristics have emerged as a practical and highly effective alternative to traditional exact methods in mixed-integer optimization. Their ability to strike a favorable balance between solution quality and computational time has made them the preferred choice for tackling complex problems and large instances. In this paper, we focus on the Generalized Discrete Cost Multicommodity Network Design Problem (GDCMNDP), a challenging network design problem. We investigate the performance of hybrid metaheuristics, specifically the Genetic Algorithm and the Non-Linear Threshold Algorithm, known for their success in diverse applications. Our proposed collaborative framework, featuring a multistage structure, harnesses the strengths of these metaheuristics. The numerical results obtained demonstrate the effectiveness of our approach in solving various test problems, highlighting its favorable performance.

Imen Mejri, Manel Grari, Safa Bhar Layeb
Isolated Handwritten Arabic Character Recognition Using Convolutional Neural Networks: An Overview

Arabic Handwriting Recognition (AHR) is a research area of great importance, given the intricacies of Arabic script. The recognition of Isolated Handwritten Arabic Characters (IHAC) is a crucial phase in AHR, and significant progress has been made in recent years, primarily due to the adoption of Convolutional Neural Networks (CNN). CNN has emerged as a powerful learning technique, dominating various computer vision-related research domains. Notably, CNNs have been extensively utilized for IHAC recognition since 2017. This paper presents an analysis of CNN-based methods employed in IHAC recognition. We delve into the advancements made in network architectures, training strategies, datasets, and results. The findings from this review emphasize the immense potential of CNN-based methods in IHAC recognition and shed light on future research directions to tackle the challenges associated with this field. Overall, CNN-based methods hold promising prospects for improving the accuracy and efficiency of IHAC recognition, which can have far-reaching applications in document analysis, text recognition, and language processing.

Mohsine El Khayati, Ismail Kich, Youssfi Elkettani
A New Approach for Quantum Phase Estimation Based Algorithms for Machine Learning

One of the greatest developments in computer science is undoubtedly quantum computing. It has demonstrated to give various benefits over the classical algorithms, particularly in the significant reduction of processing time, due to the parallelism and entanglement properties. One of the most crucial quantum computing algorithms is quantum phase estimation (QPE). It is called the eigenvalue finding module for unitary operators. It has helped to solve the order finding and the factoring problem, and to calculate the eigenvalues of unitary matrices and quantum sampling methods. In this paper, we study recent improved versions for the QPE procedure, their advantages and experimentation. We also propose a new approach for QPE based algorithms for machine learning (ML). These algorithms are the Harrow-Hassidim-Lloyd (HHL) algorithm for solving linear systems, the quantum singular value thresholding (QSVT) algorithm for matrix completion in recommender systems, and the quantum principal components analysis (QPCA) for data visualization.

Oumayma Ouedrhiri, Oumayma Banouar, Salah El Hadaj, Said Raghay
Model Risk in Financial Derivatives and The Transformative Impact of Deep Learning: A Systematic Review

This paper presents a comprehensive examination of model risk within the derivatives pricing context, specifically focusing on autocallable products. It identifies potential sources of model risk, encompassing insufficient model assumptions, calibration challenges, data availability limitations, and overfitting concerns. The paper explores diverse approaches for model risk mitigation in derivatives pricing, including robust model selection, rigorous model validation, risk sensitivity analysis, and model diversification. Moreover, it investigates the utilization of advanced machine learning techniques to alleviate model risk and discusses alternatives to tree methods, such as manifold learning algorithms and topological data analysis, for gaining deeper insights into intricate datasets and pricing relationships. The paper concludes with an analysis of autocallable notes, emphasizing the critical importance of accurately capturing correlations between underlying assets and their corresponding volatilities.

Mohammed Ahnouch, Lotfi Elaachak, Abderrahim Ghadi

Digital Twins

Integrating Syrian Cadastral Data into Digital Twins Through Accurate Determination of Transformation Parameters

Syria is a developing country adopting the local datum known as clarke1880 for all surveying and mapping activities. Currently, for all numerical maps and digital models in Syria, the use of satellite positioning technology is needed and mainly for geodetic applications using Global Navigation Satellite Systems (GNSS). The integration of Syrian cadastral data into digital twins is essential for the country construction and infrastructure projects. Therefore, the establishment of a functional relationship between the Clarke 1880 and GNSS reference datum is necessary. The problem is in one hand, the absence of accurate transformation parameters from GNSS data into the Syrian Cadastral system. In another hand, the insufficiency of geodetic network reference points in the region severely hinders attempts to make use of GNSS; some of these points have issues due to vacancies in the region while others have been damaged or lost entirely over the years.In this study, the least squares method is applied on a set of 100 points distributed across the area of Syria, to obtain the 7-transformation parameters whose global and local geocentric coordinates are known and then estimate their accuracy. These parameters were then used to transform geocentric coordinates of a set of 35 points, whose global geocentric coordinates are known, into the local geocentric coordinates. However, Syria’s national coordinate system is a projected grid coordinate, called the Syrian stereographic system (cadastral system), and thus the geocentric coordinates (X, Y, Z) on the local datum are not applicable. There is therefore the need to obtain these coordinates in the Syrian stereographic system. Statistical study of errors was carried out as well as the calculation and evaluation of the accuracy indicators to give finally necessary recommendations to integrate Syrian cadastral system in digital twins.

Al-Kasem Shaza, Ramadan A. Al-Razzak, Jibrini Hassan
Towards Linked Building Data: A Data Framework Enabling BEM Interoperability with Extended Brick Ontology

Building Energy Modeling (BEM) aims to quantify the energy performance of buildings to help designers and architects better understand the environmental impact of their decisions. BEM comprises a large amount of valuable data such as HVAC equipment, thermal envelope, and occupant behavior. It can be applied for building energy simulation, prediction, optimization, etc. Nevertheless, it typically uses a proprietary data format depending on the BEM tool which makes it a disparate data silo and hinders its interoperability with other data sources. Therefore some cross-domain applications that require the integration of data from different sources including BEM, are not easy to implement. In this work, a data framework for BEM interoperability is proposed based on Brick ontology and its extended concepts. This data framework is viable in real-world implementation by aligning BEM data with semantic web ontology Brick, enabling its interoperability with other data sources, and moving BEM towards linked building data.

Zhiyu Zheng, Esma Yahia, Elham Farazdaghi, Rani El Meouche, Fakhreddine Ababsa, Patrick Beguery
Digital Twin for Construction Sites: Concept, Definition, Steps

In the field of construction, a digital twin can provide insight into the performance of a building or infrastructure throughout its lifecycle, from design to operation and maintenance. The Digital twin as a concept is not the birth of today, it has emerged over the past decade in industry and production. Nowadays many of the fields are developing their sectors using digital twins. In fact, we are currently witnessing a revolution in the field of digital twins, particularly in the construction sector. This paper aims to provide different definitions for the digital twin in the construction sector, and represents its advantages and limits. It also outlines the steps of digitalizing a site during construction. The process of digitizing a construction includes various steps such as data collection, modeling and visualization, requiring a cooperative effort between all stakeholders. In general, the digital twin has the potential to revolutionize the construction industry, although it requires attentive planning and execution to realize all its benefits.

Mohamad Al Omari, Mojtaba Eslahi, Rani El Meouche, Laure Ducoulombier, Laurent Guillaumat
Towards Digital Twins in Sustainable Construction: Feasibility and Challenges

The term digital twin refers to a virtual copy of a physical structure. Nowadays, many research studies conducted to develop digital twins for construction. Digital twins represent a significant opportunity to improve the efficiency and sustainability of the construction industry. In this paper, we aim to verify the challenges and feasibility of implementing digital twins in sustainable construction. Here we have concentrated on two major factors i.e. the applications of digital twins in climate change and reducing CO2 emissions. We believe that the use of digital twins in sustainable construction offers great potential or transforming the construction industry towards more sustainable approaches. Using models and simulations of construction projects in advance of construction, as well as the possibility of monitoring construction projects over their lifecycle, digital twins allow stakeholders to make informed decisions, while reducing errors, costs and waste and optimizing building performance. Nevertheless, there are still challenges ahead, such as the development of data interoperability and protocols, as well as the improvement of modeling and simulation tools, to enhance the use of digital twins in sustainable construction.

Mojtaba Eslahi, Elham Farazdaghi, Rani El Meouche
Digital Twin Architectures for Railway Infrastructure

The importance of providing digital twins for infrastructure arises from the urgent necessity to enhance the efficiency, resilience, and sustainability of infrastructure systems. A digital twin (DT) refers to a digital replica of physical assets, processes and systems, it brings the opportunity to turn passive infrastructure assets into cyber-physical systems. Despite the increasing trend, the transportation industry lacks reviews on DTs, especially in the area of railway systems. This article shows several opportunities and advantages of DTs, focusing on the use of DTs in the railway sector. It also provides a review of existing DT architectures. The study highlights the importance of acquiring a systems perspective when designing DTs today to facilitate the development of interoperable systems of systems in the future. To this end, we present a conceptual architecture of railway subsystems integrating seamlessly digital models and data into a holistic large-scale railway infrastructure system.

Maryem Bouali, Muhammad Ali Sammuneh, Rani El Meouche, Fakhreddine Ababsa, Bahar Salavati, Flavien Viguier
Seismic Digital Twin of the Dumanoir Earth Dam

A digital twin in seismology is a digital model of a real seismic system, used to simulate the behavior of seismic waves in this system. This simulation allows us to test hypotheses and predict hypothetical seismic responses. This article describes the different parts and challenges to create a seismic digital twin of the dumanoir dam in Guadeloupe.

Mohamad Ali Noureddine, Florent De Martin, Rani El Meouche, Muhammad Ali Sammuneh, Fakhreddine Ababsa, Mickael Beaufils
Digital Twin Base Model Study by Means of UAV Photogrammetry for Library of Gebze Technical University

With the smart city concept, the requirement for visual presentation has increased to help decision-makers or users to properly organize, plan, and simulate large-scale and complex smart city scenarios. Therefore, the requirement of three-dimensional (3D) modeling of any structure for smart city applications has emerged and the concept of digital twin has gained importance. Unmanned Aerial Vehicle (UAV) photogrammetry method is often preferred to create a 3D model because it provides time and cost savings. This study aims to produce a high-quality 3D model of a library building located at the Gebze Technical University. For this aim, GeoMax Zoom25 and DJI Phantom 3 Advance were employed for conservative measures and acquiring UAV image process respectively. 368 images were taken and the acquired images were processed in Agisoft Metashape software. After completing the processes, the total root mean square error was calculated as 1.28 cm. The indoor solid model of the building was also produced employing conventional surveying techniques. As a result of the study, it was found that a high-accurate 3D model for integration into smart city applications can be produced based on the UAV photogrammetric method.

Bahadir Ergun, Cumhur Sahin, Furkan Bilucan
Leveraging Diverse Data Sources for ESTP Campus Digital Twin Development: Methodology and Implementation

In recent years, the concept of a digital twin has gained significant attention as a powerful tool for managing and optimizing complex systems. A digital twin is a virtual representation that mirrors the physical characteristics, behavior, and dynamics of a real-world system [1]. By integrating diverse data sources, such as Internet of Things (IoT) sensors, Geographic Information Systems (GIS), building management systems, and individual experience, a digital twin provides an accurate and comprehensive digital replica of the system [2]. This paper aims to present a methodology and implementation approach for developing a campus digital twin by leveraging diverse data sources. Our research addresses the challenge of integrating disparate data streams to create a holistic digital representation of the campus environment. By collecting and integrating data from various sources, we construct a digital twin that involves representative information, enabling a comprehensive understanding of the campus’s behavior and facilitating data-driven decision-making.

Saffa Mansour, Rita Sassine, Stéphanie Guibert

3D Models and Computer Vision

Road Traffic Noise Pollution Mitigation Strategies Based on 3D Tree Modelling and Visualisation

Mitigating noise pollution from road traffic is crucial in urban areas, and an effective solution is the use of trees as noise barriers in high-risk noise areas. This study proposes the incorporation of 3D noise visualisation within 3D building models to accurately visualise and quantify road traffic noise levels on building facades. Traffic noise propagates in all directions, necessitating the need for 3D visualisation. A selected tree belt in front of a three-storey building serves as the investigation site, where simulations compare noise levels along the facades with and without tree absorption. An equation is formulated to quantify road traffic noise levels and tree noise absorption. The study emphasizes the vital role of tree canopy in noise absorption, with properties such as leaf surface area, tree depth, and leaf frequency absorption being instrumental in determining the extent of noise reduction. To determine tree canopy properties, a LiDAR survey is employed, and convex hull visualisation of 3D trees is utilized to extract relevant properties. The study's results reveal that trees can absorb approximately 3 dB of noise levels. Overall, this research demonstrates the effectiveness of using trees as noise barriers and showcases the importance of 3D noise visualisation in understanding and mitigating road traffic noise on building facades.

Nevil Wickramathilaka, Uznir Ujang, Suhaibah Azri
Exploring Google Earth Engine Platform for Satellite Image Classification Using Machine Learning Algorithms

Google Earth Engine is a geospatial data processing platform that runs in the cloud. It offers free access to massive amounts of satellite data as well as unlimited computing power to monitor, visualize, and analyze environmental features on petabyte scale. The capability of this platform to support diverse approaches for land use and land cover (LULC) classification using both pixel based and object-oriented methods has been made possible through the provision of a variety of machine learning algorithms. Earth observation data have proven to be a valuable resource of quantitative information more consistent in time and space than traditional ground surveys. They offer numerous opportunities for mapping and monitoring urban areas, as well as a variety of physical, climatic, and socioeconomic data to support urban planning and decision-making. We used Landsat 8 satellite data to perform supervised classification in this paper, and we used three advanced machine learning methods Support Vector Machine (SVM), Random Forest (RF), and Minimum Distance (MD) to classify water areas, built-up areas, cultivated areas, sandy areas, barren areas, and forest areas on Moroccan territory. The classification results are displayed using a set of accuracy indicators that includes overall accuracy (OA) and the Kappa coefficient. We obtained 0.93 as a better accuracy for the MD algorithm, however, the worst accuracy result is 0.74 for the SVM algorithm.

Hafsa Ouchra, Abdessamad Belangour, Allae Erraissi
A Review of 3D Indoor Positioning and Navigation in Geographic Information Systems

Geographic information system (GIS) has become widespread with the management of geographically-based information, constituting a large part of the new data obtained and produced today in electronic platform. Three-dimensional (3D) models provide an overview of the more rapid establishment of new ideas that a two-dimensional (2D) visualization cannot present. Thus, 3D GIS demands data to be comprehensive and continious. 3D indoor positioning and 3D navigation are technologies that provide the ability to position and navigate indoors using these systems. The 3D navigation system is a field that can show covering objects in 3D along the route to the goal when precision is involved. In this study, navigation and indoor positioning studies in 3D GIS applications are mentioned, and particularly cutoff edge studies are included. It contains research on the use and benefits of 3D indoor modeling and navigation techniques. The literature examined for the study were evaluated, and it was concluded that 3D indoor positioning and navigation operations were generally performed with algorithms such as map-matching, location-based service, or ML/DL models. We comprehensively review positioning and navigation techniques in the 3D GIS field by combining existing studies, findings, and different approaches.

Buse Yaren Kazangirler, Ismail Rakip Karas, Caner Ozcan
Enhancing Smart City Asset Management: Integrating Versioning and Asset Lifecycle for 3D Assets Management

The management of assets in smart cities, particularly 3D assets, poses significant challenges due to their dynamic nature and the need for regular maintenance. Traditional approaches to asset management struggle to keep up with the evolving nature of smart city assets, hindering efficient planning, maintenance, and optimisation. To address these issues, this paper explores the integration of versioning and asset lifecycle management as a potential solution for effectively managing smart city assets. Versioning provides a means of tracking changes, maintaining documentation, and ensuring traceability and compliance. Its temporal information supports decision making, optimal planning, and resource allocation while enabling collaborative work among stakeholders. This study presents a conceptual framework for integrating versioning and asset lifecycle management in smart city 3D asset management processes, offering practical insights and guidance. By combining versioning with asset lifecycle management, smart cities can increase asset longevity, reduce downtime, enhance operational efficiency, and optimise smart city management processes. The ultimate goal is to enhance the efficiency, sustainability, and decision-making processes of smart city 3D asset management. This, in turn, leads to optimised urban planning, improved resource allocation, and improved performance of smart city infrastructure. With the increasing importance of smart cities and the critical role of 3D assets in urban planning and decision making, the integration of versioning and asset lifecycle management offers promising opportunities to achieve resilient and sustainable smart city infrastructure in the digital age. Nevertheless, there is a lot of complexity to be considered before implementing the proposed framework for the real case study, and a set of rules needs to be enforced when managing the model.

Nabila Husna Idris, Suhaibah Azri, Uznir Ujang
Image Transformation Approaches for Occupancy Detection: A Comprehensive Analysis

Using occupancy information in building management can help save energy and maintain user comfort, which is particularly important as energy becomes scarce and people rely more on appliances. While camera-based occupancy detection is widely adopted due to its efficacy, it also brings to the forefront a range of privacy-related issues that merit consideration. Therefore, this study proposes a method that uses environmental data to identify occupancy patterns. The technique converts time-series data into images to improve feature extraction and enhance the accuracy of occupancy detection. Three image transformation techniques are compared in the study, and the grayscale approach achieved the highest accuracy of 98.09%. In contrast, the Gramian Angular Summation Fields (GASF) along with the Gramian Angular Difference Fields (GADF) approaches had lower but still reasonable accuracy levels of 97.38% and 97.64%, respectively. The required training time for all three techniques was similar. These results suggest that the proposed grayscale approach is a suitable and efficient method for transforming images and detecting binary occupancy data.

Aya N. Sayed, Faycal Bensaali, Yassine Himeur, Mahdi Houchati
Low-Cost Global Navigation Satellite System for Drone Photogrammetry Projects

This work is part of a thesis research at IRC. The subject is the follow up of the evolution of an earthen dike subjected to the swell, its close environment as well as its maintenance by doing several data collection missions in two years. One type of data retrieved so far is the imagery of the dike taken by drone. The goal is to determine the shape of the dike, to observe the vegetation cover and to estimate the shallow bathymetry near the dike. A low cost GNSS receiver is used to determine the ground control points coordinates using the corrections of Centipede Network stations in the area. The coordinates were obtained in RTK-fixed mode. Two kinds of GNSS antennas (multiband and dual band) were used to determine the GCP’s (Ground Control Points) and CP’s (Checkpoints), we compare the accuracy of the digital models generated with both GCP’s sets for two collections of images. The best results were obtained with the multiband GNSS antenna for the collection of images with higher overlap ratio.

Muhammad Ali Sammuneh, Alisson Villca Fuentes, Adrien Poupardin, Philippe Sergent, Jena Jeong
3D Spatio-Temporal Data Model for Strata Management

The proliferation of multilevel (vertical) constructions has been a direct response to the scarcity of land available for housing and infrastructure development in urban regions. The increasing number of high-rise buildings, including multi-storey houses and apartments, has become widespread in Malaysia. In the realm of the housing sector, the implementation of proactive management is of utmost importance as it facilitates the preservation of individuals’ daily routines, while concurrently augmenting the efficiency of activities carried out in and around their livings. Inadequate strata management caused a multitude of issues for owners. In contrast to traditional housing, residential strata require a workable framework for proactive management. Traditional data models operate under the premise that data are valid solely in the current moment or at a designated point in time. When alterations, removals, or additions are made to the data contained in a particular database, the state of the database is changed to reflect the revised information. The primary objective of this study is to develop a data model for 3D strata management that takes into account both spatial and temporal considerations. The model includes all the spatial and temporal elements of 3D strata management. It is essential to incorporate both spatial and temporal aspects into the architecture of the data model for efficient 3D strata management and to achieve the goal of proactive strata management.

U. Mehmood, U. Ujang, S. Azri, T. L. Choon
Investigating Wind-Driven Rain Effects on Buildings with 3D City Building Models: An Analysis of Building Complexity Using Computational Fluid Dynamics

Wind-driven rain (WDR) has significant implications for building performance, including hygrothermal performance, interior damage, and structural cracking. This study focusses on investigating the effects of WDR on buildings using 3D city building models. The utilization of Computational Fluid Dynamics (CFD) allows an accurate computation of the wetting area, facilitating building condition assessment. To address the need for improved building representation, this study proposes the involvement of the building model based on the city building modelling standard. Two building models of different levels of detail (LoD) are employed, namely LoD1.3 and LoD2.3. By comparing wind velocity, pattern, and wetting area, the study examines how building complexity influences the interaction with WDR. The results demonstrate that the complexity of the building significantly influences the calculated wind velocity and pattern, as different building structures and designs are represented in the models. Furthermore, variations in wind parts affect the resulting wetting area, which is closely related to the rain trajectories induced by the wind direction. By focusing on areas with structural differences in building models, the study observes that the wetting area varies according to the diversity of exposed and protected regions. Consequently, it is concluded that different LoDs can lead to different WDR outcomes. Therefore, opting for a higher LoD provides more accurate observations of the WDR. In conclusion, this study highlights the importance of investigating the effects of WDR on buildings using 3D city building models. The analysis of building complexity utilizing Computational Fluid Dynamics offers valuable insights into the impact of WDR on building performance.

Nurfairunnajiha Ridzuan, Uznir Ujang, Suhaibah Azri, Liat Choon Tan, Izham Mohd Yusoff
Wildfire Detection from Sentinel Imagery Using Convolutional Neural Network (CNN)

Wildfires are a significant threat to the environment and human life, and early detection is crucial for effective wildfire management. The aim of this research work is to develop a deep learning model using CNN method for detecting wildfire using satellite imagery. The CNN model development process involves splitting the dataset into training and testing sets, pre-processing the data using ImageDataGenerator function, building a deep learning model using Sequential function, compiling the model using Adam optimizer, categorical cross-entropy loss function, and accuracy metric, and training the model using the fit generator function. The CNN model architecture includes Conv2D, MaxPooling2D, and Dense layers.The dataset was collected from Sentinel-2 L1C, including 159 fire images and 149 non-fire images from different places in the Mediterranean region of Turkey. The CNN model was developed using the Keras library and trained for 200 epochs using the Adam optimizer. The model achieved an accuracy of 92.5% and a loss of 0.22 on the test set, outperforming existing methods for wildfire detection. The research work contributes to the field of wildfire science and management by providing a deep learning CNN detection model that can accurately predict wildfire behavior using satellite imagery. The outcomes of the research work can enable more effective and efficient wildfire mitigation efforts, improving the safety and well-being of communities affected by wildfires.

Sohaib K. M. Abujayyab, Ismail R. Karas, Javad Hashempour, E. Emircan, K. Orçun, G. Ahmet
3D Spatial Queries for High-Rise Buildings Using 3D Topology Rules

3D applications and analysis for high-rise buildings require accurate 3D representation. Besides geometrical properties, topological relationships are also important in describing adjacency, containment, and connectivity information between 3D objects. Topology rules define valid topological interactions between objects. However, 2D topology rules currently implemented in spatial databases inaccurately represent 3D topological relationships due to limited 2D connectivity. This study proposes a 36IM 3D topology rule to determine 3D topological relationships between 3D objects, specifically sub-units within a high-rise building. The 36IM tested intersections between the interiors, boundaries, and exteriors of objects, whereby the highest dimension of the intersections is entered into a 3 × 3 intersection matrix. In turn, the topological relationship is determined based on logical conditions that the intersection matrix satisfies. The 36IM topology rules were implemented within an Oracle spatial database where spatial queries based on topological relationships could be executed. Five topological relationship cases were tested which includes “meets (touches)”, “disjoint”, “contains” “within (inside)” and “overlaps”. These cases are most relevant to 3D sub-units within a high-rise building as the basis for more complex spatial analysis. As a result, 3D topological relationships between 3D sub-units were able to be determined without additional storage of topological schema or data structure. The intersections could also be expressed from 0D to 3D without any object decomposition. In conclusion, 3D topological relationships provide a foundation for more complex 3D spatial analysis. These analyses are crucial especially in high-rise buildings where 3D object validity, legal boundaries, and adjacent units can be difficult to determine.

Syahiirah Salleh, Uznir Ujang, Suhaibah Azri, Tan Liat Choon

Smart Learning Systems

Data Analysis and Machine Learning for MOOC Optimization

Our study analyses learning data in the context of massive open online courses (MOOCs) that are growing in popularity, but their effectiveness and learner completion rate are often criticized. Our goal is to improve students’ completion rate by analyzing their behaviors and level of engagement. Log data generated by students’ interactions with didactic activities are used to predict student dropout, and thus improve pedagogical quality. Different machine learning approaches, artificial neural networks, SVMs, and decision trees are used for this analysis. The use of artificial intelligence models makes it possible to personalize the learning experience, detect anomalies, and take appropriate action. The three classification models used show a high accuracy, reaching almost 99% and the mean square error is very low.

El Ghali Mohamed, Atouf Issam, Talea Mohamed
Using Machine Learning to Enhance Personality Prediction in Education

In recent years, there has been an increase in interest in using machine learning (ML) techniques for educational proposals and psychological science due to ML’s effective role in improving educational system services and academic performance. ML makes the learning process more effective, personalized, and accurate. Through ML, we can discover relevant and innovative uses in the education sector, for example, adaptive learning, virtual reality, learning styles, fraud detection, analyzing success indicators, reducing school failure, smart tutoring, and smart academic orientation. This paper systematically presents a comprehensive literature review of existing personality recognition techniques from a psychological and computational perspective. Specifically, the research covered in this paper concerns, on the one hand, the feasibility of using personality traits as indicators of educational success and, on the other hand, the classification of learners according to personality type to determine the learner’s learning style and the appropriate academic orientation using ML techniques.

Hicham El Mrabet, Mohammed Amine El Mrabet, Khalid El Makkaoui, Abdelaziz Ait Moussa, Mohammed Blej
Smart Education in the IoT: Issues, Architecture, and Challenges

The integration of information and communication technologies (ICT) and the continual introduction of novel technologies into institutional learning have led to the emergence of smart education as a prevalent characteristic in the field of education. The Internet of Things (IoT) is a novel technological paradigm that involves the interconnection of objects that are outfitted with sensors, actuators, and processors, enabling them to communicate with one another and perform a purposeful function. The development of IoT applications poses a significant challenge; however, the implementation of a reference architecture can offer a viable solution for efficient design processes in the realm of smart education. This manuscript describes the evolution of intelligent education in the Internet of Things through the implementation of diverse methodologies. Additionally, it introduces a comprehensive approach that addresses the prerequisites and frameworks necessary for the realization of intelligent education, and it outlines the advantages and benefits that smart education may provide, as well as the challenges that it encounters.

Ahmed Srhir, Tomader Mazri, Mohammed Benbrahim
Enhancing Book Recommendations on GoodReads: A Data Mining Approach Based Random Forest Classification

With the rise of technology, new ways of finding books have emerged beyond traditional bookstores. Websites like allow readers to share their book reviews and ratings. This study uses the data from GoodReads to find the best way to suggest books to readers. Employing data mining classification, four methods - Random Forest, Naive Bayes, K-Nearest Neighbor, and Support Vector Classifier - were examined. Performance evaluation was conducted using accuracy, F-measure, recall, and precision metrics derived from the confusion matrix. Interestingly, the Random Forest algorithm stood out with remarkable results. It achieved 99.91% accuracy, 100% precision, 92% recall, a 95% F1-score, and a slight 0.09 average error. These impressive outcomes highlight the algorithm’s effectiveness in predicting user preferences and offering personalized book recommendations. Additionally, the study compared the Random Forest approach with the baseline methods, showing its clear superiority. This research showcases the promising potential of Random Forest in improving the GoodReads book recommendation system. Using the random forest classifier proved effective in predicting user preferences and generating relevant book recommendations, offering a promising approach to enhance personalized reading experiences, fitting well with changing reading habits in the digital era.

Sajida Mhammedi, Hakim El Massari, Noreddine Gherabi, Mohamed Amnai
Reinforcement Learning Algorithms and Their Applications in Education Field: A Systematic Review

Reinforcement Learning, a sub-field of Artificial Intelligence, has attracted considerable interest and achieved notable achievements across various domains, spanning from robotics to game playing and education. In this paper, a review of the literature regarding the implementation of Reinforcement Learning algorithms in the field of education is presented. Various studies that have employed Reinforcement Learning techniques to address different educational challenges, such as personalized learning, adaptive tutoring systems and intelligent assessment and feedback, are mentioned.

Hafsa Gharbi, Lotfi Elaachak, Abdelhadi Fennan
Machine Reading Comprehension for the Holy Quran: A Comparative Study

Question Answering (QA) has become a popular topic of research in the Natural Language Processing (NLP) community in recent years. This means that researchers and enthusiasts in the field of NLP have been actively working on developing models and improving existing ones to better answer questions. However, there are fewer studies on Arabic QA compared to other languages, and even fewer on QA for the Quran. BERT is a deep neural network model that has outperformed other models on the SQuAD benchmark. BERT is known for its ability to understand contextual information and provide accurate answers. Therefore, it is a promising model for Quranic QA. In this paper, we will abord to a comparative study of different models based on BERT and used by researchers in the religious field of MRC more precisely the Holy Quran.

Souhaila Reggad, Abderrahim Ghadi, Lotfi El Aachak, Amina Samih
Innovations in Smart Cities Applications Volume 7
Mohamed Ben Ahmed
Anouar Abdelhakim Boudhir
Rani El Meouche
İsmail Rakıp Karaș
Copyright Year
Electronic ISBN
Print ISBN

Premium Partner