HAL Portal UTC Université de Technologie de Compiègne
Not a member yet
11652 research outputs found
Sort by
DELTA: Dense Depth from Events and LiDAR using Transformer's Attention
No full textInternational audienceEvent cameras and LiDARs provide complementary yet distinct data: respectively, asynchronous detections of changes in lighting versus sparse but accurate depth information at a fixed rate. To this day, few works have explored the combination of these two modalities. In this article, we propose a novel neural-network-based method for fusing event and LiDAR data in order to estimate dense depth maps. Our architecture, DELTA, exploits the concepts of self- and cross-attention to model the spatial and temporal relations within and between the event and LiDAR data. Following a thorough evaluation, we demonstrate that DELTA sets a new state of the art in the event-based depth estimation problem, and that it is able to reduce the errors up to four times for close ranges compared to the previous SOTA
Enhancing Context‐Aware Recommender Systems Through Deep Feature Interaction Learning
No full textInternational audienceABSTRACT In the domain of context‐aware recommender systems, understanding and leveraging feature interactions is crucial for enhancing recommendation quality. Feature interactions delve into the complex interdependencies among user characteristics, item attributes, and contextual factors like time and location. Traditional models often struggle to effectively combine these diverse features, potentially leading to suboptimal recommendations. To tackle this issue, we propose enhancing context‐aware recommender systems through deep feature interaction learning. Our model, which combines BiLSTM and Hybrid Attention mechanisms, offers a sophisticated architecture designed to exploit deep feature interactions effectively. This approach ensures that our system captures essential contextual dynamics, thereby improving the effectiveness of the recommendation process. Experimental results across multiple datasets validate the efficacy of our approach, showing significant improvements in key metrics such as and compared to traditional and contemporary models. These achievements underscore our model's ability to deliver nuanced and adaptively tailored recommendations, marking a valuable contribution to the field of recommender systems
Crack depth measurement and 3D reconstruction for concrete structures: A review
No full textInternational audienceCrack depth and three-dimensional (3D) morphology are critical for assessing durability and capacity in concrete bridges, yet routine inspections remain surface-focused. This recent review synthesizes advances in non-destructive evaluation (NDE) for quantitative crack depth and 3D reconstruction, emphasizing concrete material and bridge-scale applications. Methods are grouped by sensing physics: Ultrasonics, Ground-penetrating radar/electromagnetics, and Optical/Thermal approches. A staged workflow emerges: fast noncontact screening, volumetric localization, and defensible depth confirmation, with optional 3D reconstruction where morphology drives decisions. We highlight integration into finite-element (FE) models via surface meshes, volumetric fields, and scalar depths, all with uncertainty propagation. Remaining challenges include calibration, reinforcement effects, environmental variability, and lack of bridge-scale ground truth. Standardized benchmarks, multimodal fusion, and automation are identified as key levers for advancing from pilots to routine practice.</div
Toward Secure and Safe Railway Operations: Embedding Safety in Cybersecurity Frameworks
No full textInternational audienceThe digital transformation of railway systems is enhancing automation, connectivity, and operational efficiency. However, it also introduces complex cybersecurity risks that can propagate into safety-critical failures. Existing frameworks often treat cybersecurity and safety as separate domains, limiting their ability to address the interdependent threats facing modern railway infrastructures.This paper presents a novel, integrated risk management framework that explicitly links cybersecurity threats to their potential safety consequences. Our approach is attacker-oriented, considering adversarial capabilities and intent, and safety-informed, meaning the security analysis is guided by safety-related principles and constraints to ensure the protection of safety-critical functions.The framework is grounded in, and harmonized with, international standards including ISO/IEC 27001, ISO/SAE 21434, and EN 50126, ensuring methodological consistency with current industry practices.To demonstrate the framework’s practical applicability, we conduct a detailed case study on the ERTMS/ETCS Level 1 signaling system. Using a dynamic driving emulator under realistic operational conditions, we identify key vulnerabilities, assess their exploitability by plausible attackers, and evaluate their potential to compromise safety integrity
Gaussian Process Regression for Texture Analysis: Reconstruction and Homogenization
No full textInternational audienc
Scalar-on-Function Mode Estimation Using Entropy and Ergodic Properties of Functional Time Series Data
No full textInternational audienceIn this paper, we investigate the recursive L1 estimator of the conditional mode when the input variable takes values in a pseudo-metric space. The new proposed estimator is constructed under an ergodicity assumption, which provides a robust alternative to the standard mixing processes in various practical settings. The particular interest of this contribution arises from the difficulty in incorporating the mathematical properties of a functional mixing process. In contrast, ergodicity is characterized by the Kolmogorov–Sinai entropy, which measures the dynamics, the sparsity, and the microscopic fluctuations of the functional process. Using an observation sampled from ergodic functional time series (fts), we establish the asymptotic properties of this estimator. In particular, we derive its convergence rate and show Borel–Cantelli (BC) consistency. The general expression for the convergence rate is then specialized to several notable scenarios, including the independence case, the classical kernel method, and the vector-valued case. Finally, numerical experiments on both simulated and real-world datasets demonstrate the superiority of the L1-recursive estimator compared to existing competitors
Enhancement of cell disruption and selective recovery of intracellular components from selenium-enriched Konjac fly powder assisted by pulsed electric fields
No full textInternational audienceThis study investigated the effect of pulsed electric field (PEF, E = 0–6 kV/cm, t = 0–5.65 ms) on cell disruption and selective extraction of intracellular components from selenium-enriched Konjac fly powder, in comparison with high pressure homogenization (HPH). The intracellular components extractability and cell disruption degree were evaluated by extraction indexes and cell disruption index, respectively. Results showed that PEF significantly improved cell disruption and components release. The extraction indexes increased with higher E and t. The extractability order was ionic components > carbohydrates > proteins, while the maximum selenium level was obtained at 1.13 ms. Compared to HPH, PEF favored carbohydrates extraction over proteins. For example, at 3 kJ/g, PEF gave a selectivity index (S) ≈ 3.0 (4 kV/cm) and ≈ 4.2 (6 kV/cm), while HPH gave S ≈ 2.3. Non-linear relationships between extraction indexes and cell disruption index reflected PEF differential effects on cell membranes and walls
A Memory-Enhanced Greedy Randomized Adaptive Search Procedure for the Multi-Pickup and Delivery Problem with Time Windows
No full textThis study investigates the Multi-Pickup and Delivery Problem with Time Win-dows (MPDPTW), an advanced variant of the traditional Pickup and DeliveryProblem (PDP), where each delivery is associated with multiple pickup loca-tions. This problem commonly arises in real-life scenarios such as urban lo-gistics, retail distribution, and e-commerce, where multiple pickups must beconsolidated into single deliveries, requiring efficient routing and scheduling so-lutions. The MPDPTW introduces additional complexities, including the needto respect time window constraints and precedence relations between multiplepickups and their corresponding delivery.To address this problem, we propose an optimization framework that com-bines adaptive constructive search, long-term memory, and route recombinationstrategies within an integrated architecture. The constructive phase is basedon a Greedy Randomized Adaptive Search Procedure designed to explore thesearch space effectively. Its performance is reinforced by a long-term memorythat stores feasible routes and infeasible subsets, reducing redundant evalua-tions. Finally, a post-optimization stage strategically recombines routes fromhigh-quality solutions to further minimize the total travel cost and enhance thebest-found solutions.A series of computational experiments was conducted on standard bench-mark instances to evaluate the performance of our approach. The analysisconfirms that our method retains all the best-known solutions reported in theliterature. Furthermore, it achieves an improvement of 0.41% over the best-known solutions, highlighting its potential to enhance both solution quality andcomputational efficiency
Evaluation of the efficiency of multilayer biofilters for organic pollutants and nutrient removal from raw, diluted, and pretreated olive mill wastewater
No full textInternational audienceThis study aimed to assess the efficiency of multilayer biofiltration systems for the removal of organic matter and nutrients from raw, diluted, and pretreated olive mill wastewater (OMW) by coagulation -flocculation using lime. The experimental setup consisted of four PVC columns. The first column, serving as the control column, was filled with sand only and fed with raw OMW. The remaining columns were packed with multilayer biofilters composed of 70 % sand, 20 % oyster shells, and 10 % sawdust. Each of these columns was fed with raw, diluted with domestic wastewater (1:1), and pretreated OMW. Percolation through the columns resulted in an increase in both electrical conductivity and pH values. A high removal of total COD was observed in the column fed with diluted OMW (62.75 %). Regarding dissolved COD, high removal was achieved in the control column (67.84 %), as well as in the columns fed with raw OMW (51.34 %), diluted OMW (78.61 %), and pretreated OMW (51.08 %). Phenolic compound removal also reached high percentages in the column fed with diluted OMW (64.82 %). Concerning nutrient removal efficiency, substantial removal of orthophosphate (PO₄3−), total phosphorus (TP), and ammonium (NH₄+) was achieved in the control column (58.83 % – 62.28 % – 35.12 %), the raw OMW column (47.97 % – 50.96 % – 40.07 %), and the diluted OMW column (59.44 % – 59.74 % – 49.67 %), respectively. Based on these findings, the use of sand and sand mixed with substrates, combined with diluted OMW, appears to be an effective approach for the removal of total COD (62.75 %), dissolved COD (78.61 %), phenolic compounds (64.82 %), orthophosphate (PO₄3−) (59.44 %), total phosphorus (TP) (59.74 %) and ammonium (NH₄+) (49.67 %) from OMW. Based on this preliminary study, the treated OMW may be considered for agricultural reuse, particularly in restricted irrigation applications, as long as the residual salinity and organic load do not pose risks to soil quality or plant growth
Slope of Tangent (SOT): A New Automatic Method for Segmenting Uterine Muscular Contraction Bursts from the Electrohysterogram
No full textInternational audienceThe use of electrohysterogram (EHG) - uterine muscular activity signals - along with artificial intelligence models would help in the prediction of preterm delivery and thereby save the lives of many early-delivered infants through the necessary early medical care. The informative portions in the EHG recording which could aid in the prediction of this risky delivery are the ones related to the uterine muscular contractions. Hence, the more we perform a precise segmentation of these contraction signals, the more valuable and discriminative are the data provided to the AI models, which will in turn reflect a better learning process and prediction outcome. This paper presents a new algorithm called Slope of Tangent (SOT) for an enhanced segmentation of uterine muscular contraction signals. The method is further compared to the up-to-date uterine contraction automatic segmentation methods. The results showed that the method allowed a higher number of full detections (FD=90) and partial detections (PD=263) in comparison to the wavelet H2 nonlinear correlation method (wavh2) (FD=42 and PD=145) and the sample entropy method (FD=13 and PD=138). Further work should be done in order to reduce the number of false detections given by the SOT method. In addition, the method should be further validated on open-source databases