This study evaluates the performance of selected machine learning methods, Maximum Likelihood (MLC), Random Forest (RF), Extreme Gradient Boosting (XGB), Support Vector Machine (SVM), and Artificial Neural Networks (ANN), for land use/land cover (LULC) classification using Sentinel-2 satellite imagery. Each algorithm was tested across multiple classification scenarios that systematically varied both training sample size and training sample quality. To emulate realistic imperfections in reference data, controlled levels of label noise were introduced into the training set, and the resulting changes in classification performance were analysed. In addition, each model’s susceptibility to overfitting was assessed by comparing performance on training data and independent test data. To reduce the influence of a single random draw of training and testing pixels, all sampling-based experiments were repeated 10 times using different random seed values, and the reported results were summarized using repeated-run statistics. This design enabled a comparable assessment of how classification accuracy, stability, and generalization depend on dataset size and quality. The results indicate that SVM was the most consistent and reliable method across the tested scenarios, achieving high classification accuracy over a wide range of training sample sizes and maintaining strong robustness under noisy labels. Compared with the other methods, SVM showed lower sensitivity to degraded training data and a smaller tendency to overfit, which makes it a strong baseline choice when reference data are limited or imperfect.
This study aims to couple satellite data with precise geometric models. The performance assessment of the method is based on the dataset collected at a part of the vertical base geodetic control network in the Podkarpackie province in Poland. We used baselines determined in two 12-hour-long static Global Navigation Satellite System (GNSS) measurement sessions and reference coordinates of vertical base points of the geodetic control network. Baselines obtained from satellite measurements were used for the fusion, considering the height anomalies taken from the corresponding official geoid model PL-geoid2021. Proprietary solutions played a key role in satellite measurements by ensuring the correct performance of long-term measurements without requiring the manual measurement of the height of GNSS antennas relative to the marks of the measured geodetic control points. Two patented solutions were used for the first time in this study. The results confirmed the correctness and high applicability of the presented approach.
Accurate indoor camera localization is crucial for applications in augmented reality, robotics, and autonomous navigation. While single-image deep learning models for 6-DOF pose regression have shown competitive results on established benchmarks, their development still requires extensive data annotation and hyperparameter tuning. In this work, we investigate the combination of advanced network architectures, transfer learning, and synthetic data to improve single-image indoor pose regression. Our approach employs a ResNet50 backbone pre-trained on the Places365 dataset and further trained and evaluated on established benchmarks. To enhance the training data, synthetic images are generated from 3D BIM models using Unreal Engine, with alignment procedures ensuring accurate correspondence between synthetic and real environments. Real RGB images are preprocessed to resemble synthetic data, enabling effective cross-domain evaluation. Experiments demonstrate that both architectural design and pretraining significantly influence model performance. On the UniMelb dataset (real-to-real scenario), the model achieves 0.21 m and 0.80 degrees errors, surpassing baseline accuracy. We also present cross-validation and synthetic-to-synthetic experiments, providing insights into factors affecting performance and interactions between architecture, pretraining, and dataset characteristics.
This study presents a comparative analysis of two three-dimensional reconstruction models of Powder Magazine no. 2 of the 19h-century Cracow Fortress, developed using terrestrial laser scanning (TLS) and close-range digital photogrammetry to assess the suitability of both methods for the documentation and reconstruction of heritage structures, as well as to evaluate their geometric accuracy and visual quality. The investigated structure was scanned from four positions using a Leica ScanStation P40 and photographed using a Canon EOS 400D camera. Based on the acquired point clouds, two 3D models were generated: a TLS-based model in Autodesk Revit and a photogrammetric model in Trimble SketchUp, both at Level of Detail (LOD) 200. Accuracy assessment was conducted in CloudCompare using the Cloud-to-Cloud (C2C) method, supplemented by visual and historical comparisons. The results show high geometric consistency between the datasets (with deviations below 0.2 m) and confirm the effectiveness and complementarity of both techniques in the documentation of cultural heritage. The developed models provide valuable material for scientific research, conservation planning, education, and digital heritage presentation.
The development of image-based approaches for 3D dense reconstruction has found wide application in the architectural documentation of cultural objects and sites. For this reason, it is crucial to understand the potential and limitations of Multi-View Stereo (MVS) algorithms used in 3D shape reconstruction. This article aims to evaluate the quality and accuracy of dense point generation for cultural heritage objects using open-source (MicMac, OpenMVS, RealityScan), commercial algorithms (Agisoft Metashape and Pix4D), and an Unmanned Aerial Vehicle (UAV) in parallel and skew configurations at two historic wooden structures. To achieve this, a workflow for data evaluation was proposed, incorporating eigenvalues and quality factors, such as planarity, roughness, normal vector variance, and cloud-to-TLS (Terrestrial Laser Scanning) cloud deviation analysis. The results demonstrate that the image-acquisition geometry strongly influences reconstruction accuracy. Parallel configurations consistently produced higher completeness, narrower deviation distributions, and improved geometric fidelity compared to skew configurations. Among commercial solutions, Agisoft Metashape and Pix4D achieved the best overall performance. Open-source algorithms yielded variable results: OpenMVS-SGM achieved competitive accuracy under optimal conditions, while the OpenMVS-Patch-Based Approach offered a balance between density and roughness.
The study addresses the problem of detecting displacements of building walls using terrestrial laser scanning (TLS) data and machine learning methods. Traditional displacement measurement techniques are often time-consuming. They are also limited in capturing the full geometry of monitored objects. In response, this research proposes a methodology based on the analysis of geometric and radiometric features extracted from point clouds. Controlled experiments were conducted with a geodetic rosette equipped with distance-measuring prisms, which were displaced in the XY plane by 4 mm, 9 mm, and 13 mm. Data were recorded with a Leica RTC360 scanner from three stations, yielding nine point clouds. Selected features describing differences between corresponding points in the reference and displaced series were used as input for neural network models. Both binary classification (displacement/non-displacement) and multi-class classification (0, 4, 9, 13 mm displacement) were performed. The results demonstrated high classification accuracy: 99.1% for binary models and 96.0% for multi-class models. Feature ranking revealed that geometric attributes, such as displacement vector length, curvature, and normal vectors, were the most relevant for model training, while color features had minor importance. The study confirmed that scanner position and incidence angle of the laser beam strongly affect classification quality. The developed procedure proved effective in detecting displacements that occur in directions parallel to the plane of building walls. The conclusions drawn from the research constitute a valuable contribution to the theory of monitoring building structures using TLS.
Contemporary building documentation increasingly relies on laser scanning technologies that provide rapid and precise spatial data acquisition. Portable LiDAR systems such as the Leica BLK360, as well as mobile devices like the iPad, offer modern tools for efficient documentation of interior spaces, including office rooms. With dedicated applications (e.g., Leica Cyclone Field 360, Leica Cyclone 3DR, MagicPlan, BIMx, AutoCAD Mobile), users can record measurements, create sketches, and generate 3D models and photographic documentation within a single digital environment. The purpose of this study is to assess the dimensional accuracy in the documentation of an office space using portable LiDAR systems. The experiments showed that, for most analysed dimensions, the results agreed within several centimetres, which is sufficient for architectural inventory work. However, the issue of rounded corners was identified, which may significantly influence measurement results depending on the distance-measurement method applied. Reliable accuracy analysis required proper mutual alignment (common georeferencing) of the scans acquired with the selected instruments. To achieve this, point cloud classification was performed to identify surfaces suitable for cloud-to-cloud alignment. A predefined AI-based classification model dedicated to indoor environments was used. The findings confirm that portable LiDAR systems significantly reduce the time required to complete inventory tasks and enable more comprehensive visualisation of interior spaces. This technology serves as an effective tool supporting the design, modernisation, and management of office environments in a digital workflow, although its accuracy-related limitations must be considered.
Many factors cause the deformation of slender structures, such as the weather conditions that significantly affect their geometry. Surveys of such structures are critical for deformation monitoring and, therefore, general safety. Modern surveying methods like electronic tacheometry (ET), Global Navigation Satellite Systems (GNSS), laser scanning (LiDAR, terrestrial (TLS) and mobile (MLS)), and Unmanned Aerial Vehicles (UAVs) provide means for accurate and effective monitoring. Integrating surveying technologies paves the way for an exhaustive approach to slender structure monitoring and provides accurate data to detect deformation and intervene early. Electricity transmission poles exemplify such structures. The poles need to be monitored regularly to ensure operating stability and safety. The article reports a survey of tubular steel pole deflection angles at various times of the year and under various weather conditions. The surveys were accompanied by temperature measurements and a solar irradiance analysis. The maximum horizontal deflection of the pole top was found to be approximately 27 cm over a single sunny day, highlighting the dynamic nature of solar influence. The analyses yielded conclusions regarding measurements of slender steel structures. Specifically, the study confirmed that solar irradiance can cause short-term deflections of the pole top by up to 27 cm in a single day, and uncertainty increases proportionally with structural dynamics. Consequently, recording precise time and weather conditions during surveys is essential to improve rapid, time-constrained measurement accuracy and the operational safety assessment of slender steel structures. This should improve the measurement accuracy and operational safety of slender steel structures.
This article provides a series of fictitious case studies as vignettes to exemplify how educators can support complex trauma impacted children in school to establish Flow. Flow contributes to effective immersion in learning, where educators can quickly recognise states and emotions which disrupt the Flow, and can lead to disengagement in learning. Through consideration of the case studies, educators can envisage solutions to challenges they face in their learning spaces, and most importantly support learners impacted by complex trauma to get the most out of their education and to reach their potential. This is more than Trauma Informed Practice; it is a social-emotional- psychological-physiological theory that underpins and supports daily educational practices with children and youth.
This study aims to investigate the conceptual knowledge levels and attitudes of classroom teachers toward Education 4.0 in the context of primary education in Hatay, Türkiye. The research sample comprised 419 classroom teachers employed in primary schools in the central districts of Hatay, Türkiye, selected using a simple random sampling technique. A quantitative research design employing a survey model was used. The “Education 4.0 Conceptual Knowledge Level and Attitude Scale” served as the data collection tool. Exploratory and confirmatory factor analyses were performed, confirming the validity and reliability of the scales. The Education 4.0 Conceptual Knowledge Level Scale comprised a single dimension with 23 items, while the Education 4.0 Attitude Scale included six dimensions with 35 items. Data were analyzed using the SPSS 24 package through independent samples t-tests, one-way ANOVA, and correlation analyses. The findings revealed that teachers’ conceptual knowledge levels regarding Education 4.0 were moderately sufficient, while their attitudes were generally at a high level. These results indicate that although teachers have positive perceptions of Education 4.0, further efforts are needed to enhance their conceptual understanding and technological readiness in educational contexts.
This paper explores a new perspective on the relationship between human intelligence and artificial intelligence (AI), challenging the traditional dualistic and competitive view. Drawing on the concept of a “third space” – inspired by postcolonial and philosophical theories such as those of Homi K. Bhabha and Donna Haraway – this study examines how co-creation can become a symbiotic process in which humans and AI enhance each other’s capabilities. By analyzing the ontological differences between data-based processing and conceptual understanding, we argue that competition between human and artificial intelligence is a false paradigm. Through concrete examples of textual and visual co-crafting, the paper highlights how AI can serve as a creative partner in generating new forms of “distributed intelligence.” However, we also critically examine the power dynamics, control mechanisms, and potential for digital colonialism that characterize current AI deployment. The paper concludes by proposing that AI should not be seen as a replacement for human intelligence but as a collaborative entity that requires democratic governance to foster positive and sustainable change.
We present the results of a study involving 175 Italian prospective secondary school special education teachers. Participants were engaged in a Workshop and were asked to design inclusive striped problems, that is, mathematical story-problems obtained by combining Zan’s Context and Question (C&Q) model with the principles of Universal Design for Learning (UDL). Participants worked on design activities in small groups with and without ChatGPT support. The aim of the research is to investigate the influence of the Workshop on the participants’ idea of inclusive mathematics and inclusive mathematics teaching, and to understand the support – positive or negative – offered by ChatGPT in designing inclusive learning activities in mathematics. The results of a qualitative analysis show how the training of special education teachers, as designers of inclusive teaching activities, is crucial to promoting inclusive mathematics education.
Facial expression synthesis in virtual environments is critical for educational applications, yet existing systems struggle to balance realism, cultural inclusion, and accessibility. This paper presents a multidimensional framework derived from a systematic review of 127 studies (2014–2024). The framework addresses three key tensions: (1) the realism-accessibility trade-off in generative models, (2) the imperative for cultural inclusion in expression datasets, and (3) the need for pedagogical grounding of expressive agents. The framework’s four dimensions—technical, pedagogical, sociocultural, and operational—offer a replicable blueprint for equitable educational tools. By shifting the evaluation focus from raw technical performance metrics (e.g., F1-score) to contextualized pedagogical utility and establishing a clear theoretical distinction from existing affective computing models, this work provides a nuanced and actionable guide for developers and educators.
This cross-sectional study explores how generative artificial intelligence (GenAI) tools influence English-as-a-foreign-language (EFL) writing among third-year secondary students in Algeria. Using a sample of 105 students and their teacher, the research combined classroom observations, a semi-structured interview and a questionnaire to assess improvements in grammar, vocabulary, sentence structure and organization, and to gauge perceptions of GenAI. Observations revealed that integrating GenAI tools into pre-writing, drafting and revision fostered collaborative learning, heightened engagement and encouraged metalinguistic awareness. The teacher interview underscored benefits such as personalized feedback and increased autonomy, but cautioned against over-reliance, highlighting the need for prompt-design training, ethical guidance and the cultivation of critical thinking. Moreover, students’ perceived impacts varied by sub-skill. GenAI also bolstered their confidence in writing. The study concludes that GenAI can enhance lexical richness and organizational skills when thoughtfully embedded into classroom practice, but it does not obviate the teacher’s role. A blended approach—balancing automated feedback with human instruction—appears most conducive to developing confident, independent writers in non-English-dominant contexts.
This paper discusses recent and up-to-date uplift rates at Polish Polar Station at Hornsund (SW Svalbard). Twenty years of continuous Global Navigation Satellite Systems (GNSS) measurements was used to infer contemporary vertical crust deformation due to viscous and elastic response to variable loads. Using environmental models good agreement of observed land uplift with numerical predictions was found. Most of the vertical change stem from Glacial Isostatic Adjustment (GIA), Little Ice Age (LIA) and Present Day IceMelting (PDIM). Significant increase of uplift was observed in the last few years from 9.0mmyr−1 up to 11.5mmyr−1. This can be attributed to increase of ice mass loss at Svalbard archipelago of additional 10 Gt yr−1. In case of GNSS site at Hornsund 80% of PDIM comes from melting of glaciers at south Spitsbergen.
Identification of the reference base in deformation analysis is a key element in obtaining the correct deformations of the monitored object. As it cannot be assumed that all the reference points are stable, a stability check must be conducted as the initial stage of the deformation analysis. Most known methods used for the identification of stable reference points cover leveling or horizontal networks. This study presents an approach using a coordinate transformation for the comparison of coordinates corresponding to two measurement epochs in 3D deformation analysis. The proposed algorithm is aimed at selecting such coordinate transformation parameters that correspond to the acceptable transformation residuals for the points of the reference base. Three robust adjustment weight functions (Kadaj, Huber, and Danish) were selected as a basis for formulating the objective function. The six 3D transformation parameters create a 6D search space. The optimization algorithm (simulated annealing) was applied for the search through the search space for desirable transformation parameters. The simulated example 3D two-epoch network was used for testing the performance of particular objective functions. The necessary parameters of the objective functions and optimization procedure were selected empirically. The test results confirmed the correctness of the adopted solution and the ability of all objective functions to detect the groups of fitting points. The objective function based on the Danish weight function appeared to be most promising due to the wide possibilities of shaping its features, which enable tailoring the objective function to the specific need.
Regular monitoring of dams is fundamental to ensuring their safety, stability, and smooth operation. However, during the measurement phase, a number of obstacles are encountered that are often difficult to predict at the planning stage of the observation campaign. In such situations, the surveyor must decide whether to continue all planned observations, which may involve a significant extension of working time. An alternative is to reduce the number of observations deliberately, provided the required accuracy and reliability of the data are ensured. This paper examines the impact of missing observations on the position errors of control network points. Using an angular-linear network as an example, a simulation was carried out by excluding observation stations selected at random from subsequent adjustments and by checking its effect on the final result.Additionally, the possibility of removing from the adjustment those observations that were considered problematic for measurement due to various types of obstacles was verified. The results of the conducted studies show that the cautious station elimination prevents network weakening. The necessary removal of problematic observations can be implemented already at the measurement stage. These findings highlight the significant trend of decreasing accuracy in determining the positions of network points. After eliminating 15% of the observations, the control network maintained high accuracy, with the resulting RMSE changing by no more than 10%.
Laser scanning systems are modern measurement techniques generating large datasets. Observations, usually collected as a point cloud, present the general results that can be visualized using specialized software. While the final effect might be impressive from a visualization point of view, it is inconvenient formodeling or extracting detailed information about, for example, terrain, buildings, engineering structures, and deformations. Therefore, data from laser scanning systems require post-processing using several methods reflecting different purposes or data processing stages: data segmentation, modeling, and filtration. Msplit estimation is one of the methods that has proved its effectiveness in laser scanning data processing and determination of terrain profiles, deformation, or building shapes. Processing the complete datasets tends to only yield often inadequate results when high-class computers are used, and it is time-consuming. Therefore, datasets tend to remain segmented. This paper explores a range of several types of segmentation methods that can be used in Msplit estimation. It presents profile determination when data cut out from the original point cloud are divided into intervals of the same length, or the sliding window algorithm is applied. In comparison, the given examples show that the latter approach can providemore reliable results. The application of the sliding window algorithm entails having to make assumptions concerning estimation parameters. The paper offers valuable guidance about both the width of the window and the slide size.
