Data Fusion Methodology Research Articles

Practical design and implementation of IoT-based occupancy monitoring systems for office buildings: A case study

This study introduces a scalable, cloud-based approach to occupancy monitoring designed to optimize HVAC operations in office buildings. It addresses the challenges of developing and implementing a multi-parameter IoT-based occupancy monitoring system by integrating various off-the-shelf sensors—CO2, infrared (IR), motion (PIR), and door status detection—into a cohesive system. Leveraging wireless LoRaWAN and novel cloud technologies, the system ensures easy installation, efficient maintenance, and robust data management. CO2-based occupancy detection models were trained using data from a reference office and validated in another office environment. Among the various models evaluated, the four best-performing ones—Decision Trees, Random Forest, LightGBM, and K-Nearest Neighbors—were selected for integration into a multi-parameter detection system. To further enhance system performance and identify optimal sensor combinations and configurations for cost-effective and accurate occupancy detection, a data fusion methodology was employed. This methodology, validated with ground-truth data from a test bed, tested the monitoring system in different office settings, ranging from single to quadruple-occupant rooms. Integration of additional parameters into the developed data fusion approach significantly improved system performance, achieving a True Positive Rate (TPR) of 95% compared to 81% with a simple baseline data fusion method. This approach also reduced false detections during unoccupied periods, as tested in multiple rooms within the studied building, thereby enhancing the system's reliability for integration into occupancy-aware HVAC control strategies.

Sensor placement for optimal aerodynamic data fusion

Aircraft design is recently evolving towards a digital twin representation that involves many heterogeneous data sources. The aerodynamic development of aircraft usually incorporates data from computational fluid dynamics simulations, wind tunnel testing, and flight tests. All of these data sources have their advantages and disadvantages, which can optimally be combined using data fusion methods. However, the quality of the data fusion result strongly depends on the experimental design, i.e. the placement of discrete sensors. Therefore, an optimized sensor placement is essential for data fusion applications, as the number of sensors is limited. This work presents a sensor placement strategy for the widely used Gappy proper orthogonal decomposition data fusion methodology. The sensor placement relies on a Bayesian formulation of the data fusion, allowing accurate error estimates. Based on the Bayesian posterior, a utility function characterizes the quality of the fused result by quantifying the expected information gain for the proper orthogonal decomposition coefficients. As the optimization of the sensor locations involves a complex combinatorial problem, we introduce an efficient genetic algorithm for this task. The method is demonstrated on a two-dimensional airfoil and the NASA Common Research Model with synthetic measurement errors. For both test cases, an optimal sensor placement results in smaller reconstruction errors than a conventional layout. The Bayesian approach leads, in most cases, to more accurate reconstructions and is more versatile than other well-established sensor placement methods. The proposed genetic algorithm finds better optima with significantly fewer function evaluations than the widely used greedy algorithms.

Gas permeability, diffusivity, and solubility in polymers: Simulation-experiment data fusion and multi-task machine learning

Machine learning (ML) models for predicting gas permeability through polymers have traditionally relied on experimental data. While these models exhibit robustness within familiar chemical domains, reliability wanes when applied to new spaces. To address this challenge, we present a multi-tiered multi-task learning framework empowered with advanced machine-crafted polymer fingerprinting algorithms and data fusion techniques. This framework combines scarce “high-fidelity” experimental data with abundant diverse “low-fidelity” simulation or synthetic data, resulting in predictive models that display a high level of generalizability across novel chemical spaces. Additionally, this multi-task scheme capitalizes on known physics and interrelated properties, such as gas diffusivity and solubility, both of which are closely tied to permeability. By amalgamating high throughput generated simulation data with available experimental data for gas permeability, diffusivity, and solubility for various gases, we construct multi-task deep learning models. These models can simultaneously predict all three properties for all gases under consideration, with markedly enhanced predictive accuracy, particularly compared to traditional models reliant solely on experimental data for a singular property. This strategy underscores the potential of coupling high-throughput classical simulations with data fusion methodologies to yield state-of-the-art property predictors, especially when experimental data for targeted properties is scarce.

Real-time vision-inertial landing navigation for fixed-wing aircraft with CFC-CKF

Vision-inertial navigation offers a promising solution for aircraft to estimate ego-motion accurately in environments devoid of Global Navigation Satellite System (GNSS). However, existing approaches have limited adaptability for fixed-wing aircraft with high maneuverability and insufficient visual features, problems of low accuracy and subpar real-time arise. This paper introduces a novel vision-inertial heterogeneous data fusion methodology, aiming to enhance the navigation accuracy and computational efficiency of fixed-wing aircraft landing navigation. The visual front-end of the system extracts multi-scale infrared runway features and computes geo-reference runway image as observation. The infrared runway features are recognized efficiently and robustly by a lightweight end-to-end neural network from blurry infrared images, and the geo-reference runway is generated through projection of the runway’s prior geographical information and prior pose. The fusion back-end of the navigation system is the Covariance Feedback Control based Cubature Kalman Filter (CFC-CKF) framework, which tightly integrates visual observations and inertial measurements for zero-drift pose estimation and curbs the effect of inaccurate kinematic noise statistics. Finally, real flight experiments demonstrate that the algorithm can estimate the pose at a frequency of 100 Hz and fulfill the navigation accuracy requirements for high-speed landing of fixed-wing aircraft.

Establishment and Optimization of the Early Warning System for Student Academy:multimodal data fusion based on deep learning

In the realm of education, the establishment of effective Academic Early Warning Systems (AEWS) holds significant promise for improving student outcomes and institutional efficacy. This paper proposes a novel approach to AEWS, leveraging multimodal data fusion and deep learning methodologies to provide a holistic understanding of students' academic trajectories. By integrating diverse data modalities including academic performance records, student engagement metrics, and socio-economic factors, our AEWS offers a comprehensive view of students' experiences. Deep learning techniques facilitate the fusion of heterogeneous data sources, enabling the extraction of meaningful patterns and predictive insights. Through iterative refinement and validation, our system aims to deliver actionable insights for educators and administrators, fostering timely interventions and personalized support strategies. The proposed AEWS represents a significant step towards enhancing student success rates and fostering a more inclusive educational environment.

Degradation prognostics of aerial bundled cables based on multi-sensor data fusion

ABSTRACT Development of advanced health monitoring sensors and high-performance computing enabled multi-sensors information to analyse degradation in complex engineering systems/components. These sensors are often aimed to capture different physical aspects of a system. Thus, each sensor only contains partial information about the same degradation process. A significant need is to devise a contemporary data fusion method that effectively integrates independent multi-sensors measurements leading to a better prognosis of the degradation process. Unlike conventional data fusion methodologies that fuse multiple sensors’ information prior implementation of prognosis step, this paper presents a novel fusion framework based on multi-sensor prognosis data. The framework is developed using Particle Filtering (PF) technique coupled with Auto-Regressive Integrated Moving Average (ARIMA) model. The proposed framework is applied on historical Non-Destructive Testing (NDT) database generated through Ultrasonic Probe Listening and Thermal Imaging of in-service aerial bundled cables (ABCs) installed in coastal regions to evaluate cable degradation over time. Promising results of f -steps prediction scheme from fused degradation values, as compared to using individual measurement mode data, indicates the efficacy of the proposed method.

Design of image intelligent focusing system based on improved SMD function and RBF algorithm.

The utilization of digital statistical processes in images and videos can effectively tackle numerous challenges encountered in optical sensors. This research endeavors to overcome the limitations inherent in traditional focus models, particularly their inadequate accuracy. It aims to bolster the precision of real-time perception and dynamic control by employing enhanced data fusion methodologies. The ultimate objective is to facilitate information services that enable seamless interaction and profound integration between computational and physical processes within an open environment. To achieve this, an enhanced sum-modulus difference (SMD) evaluation function has been proposed. This innovation is founded on the concept of threshold value evaluation, aimed at rectifying the accuracy shortcomings of traditional focusing models. Through the computation of each gray value after threshold segmentation, the method identifies the most suitable threshold for image segmentation. This identified threshold is then applied to the focus search strategy employing the radial basis function (RBF) algorithm. Furthermore, an intelligent focusing system has been developed on the Zynq development platform, encompassing both hardware design and software program development. The test results affirm that the focusing model based on the improved SMD evaluation function rapidly identifies the peak point of the gray variance curve, ascertains the optimal focal plane position, and notably enhances the sensitivity of the focusing model.

Human-centric multimodal fusion network for robust action recognition

Skeleton-based methods have made remarkable strides in human action recognition (HAR). However, the performance of existing unimodal approaches is still limited by the lack of diverse visual features in skeleton data. Concretely, due to the absence of interaction information between individuals and objects, skeleton-based methods tend to confuse similar actions. Moreover, the view invariant property of unimodal models is susceptible to restrictions. In this work, we propose an innovative skeleton-guided multimodal data fusion methodology that transforms depth, RGB, and optical flow modalities into human-centric images (HCI) based on keypoint sequences. Building upon this foundation, we introduce a human-centric multimodal fusion network (HCMFN), which can comprehensively extract the action patterns of different modalities. Our model significantly enhances the performance of skeleton-based techniques, achieving remarkable results with rapid inference speed. Extensive experiments on two large-scale multimodal datasets, namely NTU RGB+D and NTU RGB+D 120, validate the capacity of HCMFN to bolster the robustness of skeleton-based methods in two challenging HAR tasks: (1) discriminating between actions with subtle inter-class differences, and (2) recognizing actions from varying viewpoints. Compared to state-of-the-art multimodal methods, our HCMFN achieves exciting results.

Multiparametric analysis and authentication of Argentinian vinegars from spectral sources

Ultraviolet-visible (UV-Vis) and near infrared (NIR) spectroscopies allied to chemometrics were investigated for quality control and authentication of Argentinean wine and balsamic vinegars. First, a multiparametric approach was conducted to acquire predictive models by using partial least squares regression (PLS) to quantify total acidity, volatile acidity, fixed acidity, pH and total polyphenols that are the main quality parameters used to control products. Individual UV-Vis and NIR sensors as well as merged data were assessed. Reliability models with correlation coefficients higher than 0.99 and prediction error lesser than 2.2 were acquired for the UV-Vis data. Furthermore, a classification approach was performed on wine vinegar samples to classify them according to their acetification process. At first, the data provided by each individual sensor (UV-Vis and NIR) were separately analyzed by PLS-discriminant analysis. Then, datasets were jointly analyzed by applying sequential and orthogonalized PLS coupled with linear discriminant analysis (SO-PLS-LDA). The overall accuracy of the fused model reached an optimal performance with a value of 0.92 in the prediction stage. Finally, according to the analysis proposed, this work reveals when it is proper to conduct a data fusion methodology.

A multi-model data fusion methodology for reservoir water quality based on machine learning algorithms and bayesian maximum entropy

A major concern in the management of reservoirs is water quality because of the negative consequences it has on both environment and human life. Artificial Intelligence (AI) concept produces a reliable framework to recognize complicated and non-linear correlations between input and output data. Although various machine learning (ML) algorithms in recent studies were employed to predict water quality variables, the existing literature lacks exploring the combination of these algorithms, which has the potential to significantly amplify the outcomes achieved by individual models. Thus, the current study aims to bridge this knowledge gap by evaluating the precision of Random Forest Regression (RFR), Support Vector Regression, Multilayer Perceptron (MLP), and Bayesian Maximum Entropy-based Fusion (BMEF) models to estimate such water quality variables as dissolved oxygen (DO) and chlorophyll-a (Chl-a). The comparisons were conducted in two primary stages: (1) a comparison of the outcomes of different ML algorithms with each other, and (2) comparing the ML algorithms' findings with that of the BMEF model, which considers uncertainty. These comparisons were evaluated using robust statistical measures, and, finally, to indicate the utility and efficacy of the newly introduced framework, it was efficiently utilized in Wadi Dayqah Dam, which is situated in Oman. The findings indicated that, throughout both training and testing phases, the BMEF model outperformed individual machine learning models, namely MLP, RFR, and SVR by 5%, 26%, and 10%, respectively, when R2 and Chl-a are considered as evaluation index and water quality variables, respectively. Additionally, as the individual ML models are not capable of predicting electrical conductivity and oxidation-reduction potential efficiently, the BMEF model leads to better results by R2=0.89, which outperforms MLP (R2=0.81), RFR (R2=0.79), and SVR (R2=0.62) for oxidation-reduction potential. Regarding the study limits of the present study, spatio-temporal data should be collected over a long time to increase the data frequency and reduce the uncertainty related to climate variability.

The Innovative Growth of Space Archaeology: A Brief Overview of Concepts and Approaches in Detection, Monitoring, and Promotion of the Archaeological Heritage

Space Archaeology (SA), also known as Satellite Archaeology, Satellite Remote Sensing for Archaeology, or Archaeology from Space, is part of the wider interdisciplinary field of Remote Sensing for Archaeology. The application of satellite data in archaeological investigations has proven useful for landscape observation and analysis, the detection of archaeological traces, the reconstruction and monitoring of natural and anthropic processes, and the management and promotion of archaeological heritage. During the last few decades, the increasing number of SA studies has demonstrated innovative growth in archaeological disciplines due to the significant enhancement of spatial technologies, the advancement of visual inspection and image processing techniques, the development of data fusion methodologies, and the improvement of multi-temporal analysis methods. Therefore, a broad overview of the current situation in the concepts and approaches of SA is necessary to gain greater awareness of the current potentialities and limitations of this science to better address future studies. The present work provides a review of the scientific literature by exploring the different aspects of SA and the goals achieved to date in three main application fields: Detection, monitoring, and promotion of archaeological heritage. The contributions reviewed were divided within each of these three categories by analyzing the type of SA data and methods applied. The results indicate that (1) most studies aim to detect archaeological sites compared to monitoring and promotion; (2) optical images are used more than SAR data; and (3) techniques of image enhancement and visual interpretation are applied more than other data processing methods. This brief overview allows us to consider SA as an evolving discipline, an engine for cross-skills training, as well as a promising future science that can play a key role in the frontier of sustainable development and the new space economy.

Aero-engine remaining useful life prediction method with self-adaptive multimodal data fusion and cluster-ensemble transfer regression

Remaining useful life (RUL) prediction based on multimodal sensing data is indispensable for predictive maintenance of aero-engine under cross-working conditions. Although data-driven methods have emerged as a powerful tool in RUL prediction, it is still limited in industrial applications because the majority of existing methods manually select or fuse multisensory data and ignore the inconsistency of the sensing data collected from different engines. Therefore, an intelligent RUL prediction approach is proposed for aero-engine by integrating multimodal data fusion methodology and ensemble transfer learning technology to dynamically select sensing data and make a robust RUL prediction under cross-working conditions. Specifically, a self-adaptive dynamic clustering approach is developed to select useful multimodal data into different clusters, each of which has a consistent degradation tendency. Furthermore, a cluster-ensemble transfer regression network is constructed by building multiple regressors for different clusters to predict the RUL values of aero-engine under cross-working conditions, where a multi-level feature learning strategy is provided to learn the domain-invariant temporal degradation knowledge. Comparative experiments are conducted on the NCMAPSS dataset released in 2021. The results show that the proposed method outperforms other state-of-the-art RUL prediction methods.

Individual Micron-Sized Aerosol Qualitative Analysis-Combined Raman Spectroscopy and Laser-Induced Breakdown Spectroscopy by Optical Trapping in Air.

The attribution of single particle sources of atmospheric aerosols is an essential problem in the study of air pollution. However, it is still difficult to qualitatively analyze the source of a single aerosol particle using noncontact in situ techniques. Hence, we proposed using optical trapping to combine gated Raman spectroscopy with laser-induced breakdown spectroscopy (LIBS) in a single levitated micron aerosol. The findings of the spectroscopic imaging indicated that the particle plasma formed by a single particle ablation with a pulsed laser within 7 ns deviates from the trapped particle location. The LIBS acquisition field of view was expanded using the 19-bundle fiber, which also reduces the fluctuation of a single particle signal. In addition, gated Raman was utilized to suppress the fluorescence and increase the Raman signal-to-noise ratio. Based on this, Raman can measure hard-to-ionize substances with LIBS, such as sulfates. The LIBS radical can overcome the restriction that Raman cannot detect ionic chemicals like fluoride and chloride in halogens. To test the capability of directly identifying distinctive feature compounds utilizing spectra, we detected anions using Raman spectroscopy and cations using LIBS. Four typical mineral aerosols are subjected to precise qualitative evaluations (marble, gypsum, baking soda, and activated carbon adsorbed potassium bicarbonate). To further validate the application potential for substances with indistinctive feature discrimination, we employed machine learning algorithms to conduct a qualitative analysis of the coal aerosol from ten different origin regions. Three data fusion methodologies (early fusion, intermediate fusion, and late fusion) for Raman and LIBS are implemented, respectively. The accuracy of the late fusion model prediction using StackingClassifier is higher than that of the LIBS data (66.7%) and Raman data (86.1%) models, with an average accuracy of 90.6%. This research has the potential to provide online single aerosol analysis as well as technical assistance for aerosol monitoring and early warning.

Spatiotemporal fusion of multi-source remote sensing data for estimating aboveground biomass of grassland

Accurate estimation of aboveground biomass of grasslands is key to sustainable grassland utilization. However, most satellites cannot provide high temporal and spatial resolution data. Patterns of grassland dynamics associated with variability in climate conditions across spatiotemporal scales are yet to be adequately quantified. A spatiotemporal fusion model offers the opportunity to combine the resolution advantages of different remote sensing data to achieve a high frequency and high precision monitoring of vegetation. We test a flexible spatiotemporal data fusion (FSDAF) methodology to generate synthetic normalized difference vegetation index (NDVI) data from Moderate-Resolution Imaging Spectroradiometer (MODIS) and Landsat data sets. The methodology is tested for the semi-arid grassland of the Xilin River Basin, China. Based on NDVI data fusion and field measured aboveground biomass an aboveground biomass estimation model is established for the watershed. Exploring the temporal and spatial changes of biomass and its relationship with environmental factors. The results show that: (1) The FSDAF model performs well (R2 = 0.75) and has clear textural features. (2) The established Support Vector Machine Aboveground Biomass model not only ensured the accuracy of estimation (R2 = 0.78, RMSE = 15.43 g/m2), but also generated spatiotemporal maps of biomass with higher spatial (30 m) and temporal resolution (8 days). (3) The grassland aboveground biomass in this area decreases from southeast to northwest, and the grassland biomass reaches its peak at the end of July. The average biomass of different grasslands decreases in the order of meadow grassland > typical grassland > desert grassland. (4) Aboveground biomass increased linearly with increasing water content, organic carbon and total nitrogen, and was most sensitive to soil water content. During the early growing and rapid growing period, aboveground biomass is mainly affected by both air temperature and precipitation, while the effects of temperature and human activities gradually dominate in the middle and late growing periods. This study helps to improve the spatial and temporal resolution of dynamic monitoring of grassland biomass, and provides a scientific basis for grassland protection and management in arid and semi-arid regions.

Train-Centric Communication Based Autonomous Train Control System

Driverless train control systems are becoming popular in the world. Notably, this technology has already been applied for several years. However, these kinds of Automatic Train Operation systems are still under the framework that network elements and movement authority are generated from the ground control center. Hence, the aforementioned rail systems are “automated” but not “autonomous,” and the architecture utilization and train tracking interval are limited. With the development of advanced sensors and control algorithms, the train intends to be equipped with a decision-making capability. In order to further improve train movement coordination and efficiency, an Autonomous Train Control System (ATCS) is proposed. This paper provides three main contributions: Firstly, the system structure is described, and the information following in ATCS is discussed with general operation scenarios. Secondly, data isolation is broken under ATCS, and data prediction and edge-based information fusion are applied to process the real dynamic data and estimate the train's speed and position. Thirdly, with the assistant of data fusion, model predictive and data-based control methodologies are discussed.

A Bayesian-based data fusion methodology and its application for seismic structural health monitoring of the Consoli Palace in Gubbio, Italy

Recent earthquakes have demonstrated that monumental structures located in regions characterized by high seismic hazard are particularly sensitive to damage, stimulating a growing attention to the formulation of cost-effective and long-lasting methods for damage assessment. Generally, the evaluation of a healthy or damaged state is data-driven and it can be subjected to a large amount of uncertainty. In order to associate a damage symptom to an actual structural damage, including all the uncertainties involved in the process, a Bayesian-based data fusion methodology is proposed. To this purpose, different sources of information are combined, such as dynamic structural properties extracted from monitoring data (natural frequencies and mode shapes), static response data (crack amplitudes) and visual inspections. More in depth, the proposed procedure comprises three fundamental steps: i) calibration of a finite element (FE) model, partitioned in well-thought-out macro-elements on the basis of engineering judgments and/or numerical simulations and, subsequently, construction of a tuned surrogate model (SM) considering pre-selected uncertain parameters as inputs, such as the Young's modulus, shear modulus, Poisson's ratio and mass density associated to each macro-element; ii) solve the Bayesian-based inverse problem aimed at deriving the posterior statistics of the uncertain parameters over the space of the surrogate model's classes in a computational effective manner by using dynamic data; iii) adjust the posterior distribution on the basis of the information obtained from static data and visual inspections, i.e., data fusion. The suitability of the proposed approach is demonstrated by using the monitoring data pertaining to a monumental palace, located in Gubbio (Italy) and named Consoli Palace, which has been monitored by the Authors since 2017.

Detection of superfluous in channels using data fusion with wireless sensors and fuzzy interface algorithm

Without integrating various sensor data sets, sensing information in the presence of leakage for large-scale pipeline systems is very challenging. A data fusion methodology, wherein more sensor data is merged to give relevant information, is necessary to transform the challenging process into a straightforward step-by-step operation. Ultrasonic sensors are used in stage 1 to identify any ambiguities in pipeline systems, and various sites are used to gauge the rate of leak detection. As a result, a novel model for estimating various types of gas leakage in pipeline systems is examined, put to the test, and contrasted. Five distinct scenarios are seen during the leakage testing procedure using data fusion, where the optimization is done using the fuzzy interface technique. This integration procedure detects leakage rates with high accuracy, and in every test instance, the best outcomes are obtained. Additionally, the predicted model can be used in real-time with a low failure rate of numerous sensors, with MATLAB being used to simulate the results.

Sensor Fusion with Deep Learning for Autonomous Classification and Management of Aquatic Invasive Plant Species

Recent advances in deep learning, including the development of AlexNet, Residual Network (ResNet), and transfer learning, offer unprecedented classification accuracy in the field of machine vision. A developing application of deep learning is the automated identification and management of aquatic invasive plants. Classification of submersed aquatic vegetation (SAV) presents a unique challenge, namely, the lack of a single source of sensor data that can produce robust, interpretable images across a variable range of depth, turbidity, and lighting conditions. This paper focuses on the development of a multi-sensor (RGB and hydroacoustic) classification system for SAV that is robust to environmental conditions and combines the strengths of each sensing modality. The detection of invasive Hydrilla verticillata (hydrilla) is the primary goal. Over 5000 aerial RGB and hydroacoustic images were generated from two Florida lakes via an unmanned aerial vehicle and boat-mounted sonar unit, and tagged for neural network training and evaluation. Classes included “HYDR”, containing hydrilla; “NONE”, lacking SAV, and “OTHER”, containing SAV other than hydrilla. Using a transfer learning approach, deep neural networks with the ResNet architecture were individually trained on the RGB and hydroacoustic datasets. Multiple data fusion methodologies were evaluated to ensemble the outputs of these neural networks for optimal classification accuracy. A method incorporating logic and a Monte Carlo dropout approach yielded the best overall classification accuracy (84%), with recall and precision of 84.5% and 77.5%, respectively, for the hydrilla class. The training and ensembling approaches were repeated for a DenseNet model with identical training and testing datasets. The overall classification accuracy was similar between the ResNet and DenseNet models when averaged across all approaches (1.9% higher accuracy for the ResNet vs. the DenseNet).

Data fusion of geotechnical and geophysical data for three-dimensional subsoil schematisations

Subsoil schematisations are a paramount activity for the execution of infrastructure earthworks. Currently, subsoil schematisations are laborious, and mostly based on correlation and interpolation of available geotechnical and geophysical in-situ data. Geotechnical in-situ testing tends to be accurate but merely provides local information, while geophysical investigations are often performed to image the subsurface, and provide valuable insight for areas. Yet, there is not always a clear way to reflect geophysical properties to soil schematisation. This paper presents a data fusion methodology to perform subsoil schematisation and parametrisation. This methodology makes use of available geotechnical, geological and geophysical data sets, and combines them by means of machine learning algorithms (Neural Networks and Random Forest). The data fusion method is applied on two case studies. The first case study concerns a large area of a flood defence line, and the second case study concerns a small polder area next to a dike. The accuracy of the data fusion algorithm is assessed by comparing its results against a validation data set, that has not been exposed to the data fusion algorithm. The performance of the two algorithms and of the parameters that govern each algorithm is discussed. The results show that both Neural Networks and Random Forest are suitable to perform subsoil schematisations. The analyses show that the Random Forest leads to a lower error on the validation data set. However, Random Forest fails to predict the occurrence of thin clay layers in the second case study, while Neural Networks are successful at it. The data fusion methodology shows the potential to enhance the subsoil schematisation procedure, by increasing the schematisation spatial resolution.

Vision-Cloud Data Fusion for ADAS: A Lane Change Prediction Case Study

With the rapid development of intelligent vehicles and Advanced Driver-Assistance Systems (ADAS), a new trend is that mixed levels of human driver engagements will be involved in the transportation system. Therefore, necessary visual guidance for drivers is vitally important under this situation to prevent potential risks. To advance the development of visual guidance systems, we introduce a novel vision-cloud data fusion methodology, integrating camera image and Digital Twin information from the cloud to help intelligent vehicles make better decisions. Target vehicle bounding box is drawn and matched with the help of the object detector (running on the ego-vehicle) and position information (received from the cloud). The best matching result, a 79.2% accuracy under 0.7 intersection over union threshold, is obtained with depth images served as an additional feature source. A case study on lane change prediction is conducted to show the effectiveness of the proposed data fusion methodology. In the case study, a multi-layer perceptron algorithm is proposed with modified lane change prediction approaches. Human-in-the-loop simulation results obtained from the Unity game engine reveal that the proposed model can improve highway driving performance significantly in terms of safety, comfort, and environmental sustainability.