Departamento Informática y Automática

Unified Multi-Task Learning vs. Decoupled Transformer-based Perception: A Comparative Analysis

Mon, 02 Mar 2026 00:00:00 GMT

[EN]Efficient environmental perception is a cornerstone of Advanced Driver Assistance Systems (ADAS) and autonomous driving. A persistent architectural dilemma in this domain is whether to employ unified Multi-Task Learning (MTL) frameworks, which optimize computation through shared backbones, or modular multi-model pipelines, which prioritize task-specific accuracy. This paper presents a comparative analysis of these two paradigms for joint object detection and drivable area estimation. Specifically, we evaluate YOLOPX, a representative anchor-free MTL architecture, against a decoupled multi-model system that integrates RT-DETRv2 for vehicle detection and the lightweight YOLO11n-seg for drivable area segmentation on the BDD100K benchmark under identical hardware conditions. The results show that, although the MTL YOLOPX model achieves higher throughput, the decoupled system delivers substantially better detection performance, particularly in the stricter 𝑚𝐴𝑃 50:95 metric, while preserving competitive segmentation quality and maintaining real-time latency suitable for edge deployment. These findings suggest that modular designs, rather than monolithic MTL models, can offer a more favorable balance between safety-critical detection accuracy and computational efficiency for next-generation intelligent vehicles.

Mining patient data from heterogeneous sources for decision making on administration of non invasive mechanical ventilation in intensive care units.

Wed, 01 Jan 2014 00:00:00 GMT

[EN]This paper addresses the problem of decision making regarding the administration of non invasive mechanical ventilation in intensive care units. The great number of factors to take into account, its heterogeneity and diverse origin make very difficult this process. In order to facilitate this task we propose the application of data mining methods to extract knowledge from the wide and complex information available. The aim is to find out the factors influencing the success/failure of NIMV and to predict the results in future patients. These methods have not been previously applied in this field in spite of the good results obtained in other medical areas. In this work a comparative study of different algorithms has been carried out using a wide spectrum of data obtained during 6 years about 389 patients that received treatment with NIMV. The results reveal that some multiclasifiers can be useful tools for helping physicians in the choice of the best action.

Machine Learning Methods for Mortality Prediction of Polytraumatized Patients in Intensive Care Units – Dealing with Imbalanced and High-Dimensional Data

Wed, 01 Jan 2014 00:00:00 GMT

[EN]The aim of this study is the prediction of death of polytraumatized patients based on epidemiological, clinical and health treatment variables by means of data-mining methods. The main problems to be addressed were high dimensionality and imbalanced data. Since the techniques usually used to deal with these drawbacks, as feature selection methods and sampling strategies respectively, did not provided satisfactory results, the aim of the study was to find out the data mining algorithms showing the best behavior in this kind of scenarios. The study was carried out with data from 497 patients diagnosed with severe trauma who were hospitalized in the Intensive Care Unit (ICU) of the University Hospital of Salamanca. The results of the study reveal the better behavior of multiclassifiers as compared with simple classifiers in contexts of high dimensionality and imbalanced datasets, without the need to resort to undersampling and oversampling strategies, which can lead to the loss of valuable data and overfitting problems respectively.

Computational study of the effect of forced exercise (swim stress) on the NADPH- diaphorase activity in the supraoptic nucleus

Tue, 01 Jan 2002 00:00:00 GMT

[EN]Nitric oxide synthesizing neurons of the hypothalamus show changes following a variety of experimental and pathological conditions, such as salt loading, hypotension, cholestasis, water deprivation, and several types of stress including immobilization, exposure to low temperature and changes of environment. Recently, our group has shown that forced exercise (swim stress) is an additional stressor for the nitrergic neurons located in the paraventricular nucleus of the hypothalamus. On the other hand it is currently well-known, by means of histochemical (NADPH-diaphorase), immunohistochemical and hybridization methods, that the supraoptic nucleus contains an important nitrergic population of neurons. In the present study, the effect of forced exercise on these neurons was investigated in a computational study of the NADPH-diaphorase positive population of the supraoptic nucleus. A significant increase in the number of positive neurons was observed following forced swimming, especially after 30 and 40 min. These data indicate: 1.– The influence of stress on the NADPH-diaphorase-activity in the supraoptic nucleus. 2.– The involvement of hypothalamic nitric oxide synthesizing-neurons in the response to different types of acute stressors. 3.– The excellence of a combination of morphological and computational tecniques for the detection of changes in plasticity of the hypothalamic neurons.

Predictors of the post-stroke status in the discharge from the hospital. Importance in nursing

Sun, 01 Jan 2023 00:00:00 GMT

[ES]A menudo, por parte del paciente y de la familia, se solicita a los profesionales de enfermería que predigan los factores que influyen en el estado post-ictus. Se han realizado numerosos estudios para determinar los factores que influyen en el estado neurológico post-ictus en el momento del alta hospitalaria. Sin embargo, las técnicas de aprendizaje automático no se han utilizado para este propósito. Con el objetivo de obtener reglas de asociación del pronóstico neurológico, se ha llevado a cabo un doble análisis, tanto clínico como con técnicas de aprendizaje automático, de las posibles asociaciones de factores que influyen en el estado neurológico de los pacientes post-ictus. El algoritmo Apriori detectó varias reglas de asociación con alta confianza (≥ 95%), con el siguiente patrón: En pacientes en el rango de edad de 50-80 años, la asociación de un NIHSS entre 11 y 15 puntos (NIHSS intermedio/bajo), junto con la trombectomía, conduce a la recuperación ad integrum al alta. Con la técnica de remuestreo SMOTE, se alcanzó el 100% de confianza para la asociación de NIHSS elevado (>20) y afectación de las arterias carótida y basilar, con pronóstico nefasto (exitus). Estas reglas confirman, por primera vez con aprendizaje automático, la importancia de la asociación de algunos predictores, en el pronóstico post-ictus. El conocimiento por parte de las enfermeras de estas reglas puede mejorar los resultados del ictus. Adicionalmente, el papel de la enfermería en los programas de educación sobre los factores de riesgo, y pronóstico de un ictus se torna imprescindible.; [EN]Nurses are often asked to predict factors that influence post-stroke outcome by the patient and family. Many studies have been carried out in order to determine the factors that influence the neurological status of the post-stroke patient at the moment of the discharge from the hospital. However, machine learning techniques have not been used for this purpose. Therefore, with the objective of obtaining association rules of neurological prognosis, a double analysis, both clinical and with machine learning techniques of the possible associations of factors that influence the neurological status of the post-stroke patients has been carried out. The Apriori algorithm detected several association rules with high confidence (≥ 95%), from which the following pattern: In patients in the age range of 50-80 years, the association of a NIHSS between 11 and 15 points (intermediate/low NIHSS), along with thrombectomy, leads to recovery ad integrum at discharge. With the SMOTE resampling technique, the 100% confidence was reached for the association of high NIHSS (>20) and involvement of the carotid and basilar arteries, with a dire prognosis (exitus). These rules confirm, for the first time with machine learning, the importance of the association of some predictors, in the post-stroke prognosis. The knowledge by the nurses of these association rules can successfully improve stroke outcome. In addition, the role of nurses in education programs that teach knowledge of risk factors and stroke prognosis becomes essential.

Prognostic factors associated with mortality in patients with severe trauma: From prehospital care to the Intensive Care Unit

Thu, 01 Jan 2015 00:00:00 GMT

[EN]Objective To identify factors related to mortality in adult trauma patients, analyzing the clinical, epidemiological and therapeutic characteristics at the pre-hospital levels, in the Emergency Care Department and in Intensive Care. Design A retrospective, longitudinal descriptive study was carried out. Statistical analysis was performed using SPSS, MultBiplot and data mining methodology. Setting Adult multiple trauma patients admitted to the Salamanca Hospital Complex (Spain) from 2006 to 2011. Main variables of interest Demographic variables, clinical, therapeutic and analytical data from the injury site to ICU admission. Evolution from ICU admission to hospital discharge. Results A total of 497 patients with a median age of 45.5 years were included. Males predominated (76.7%). The main causes of injury were traffic accidents (56.1%), precipitation (18.4%) and falls (11%). The factors with the strongest association to increased mortality risk (p < 0.05) were age > 65 years (OR 3.15), head injuries (OR 3.1), pupillary abnormalities (OR 113.88), level of consciousness according to the Glasgow Coma Scale ≤ 8 (OR 12.97), and serum lactate levels > 4 mmol/L (OR 9.7). Conclusions The main risk factors identified in relation to the prognosis of trauma patients are referred to the presence of head injuries. Less widely known statistical techniques such as data mining or MultBiplot also underscore the importance of other factors such as lactate concentration. Trauma registries help assess the healthcare provided, with a view to adopt measures for improvement.; [ES]Objetivo Identificar los factores relacionados con la mortalidad de los pacientes adultos politraumatizados, analizar las características clínicas, epidemiológicas y terapéuticas en los niveles prehospitalario, Servicio de Urgencias y Cuidados Intensivos. Diseño Estudio retrospectivo, longitudinal y descriptivo. Análisis estadístico a través del programa SPSS, MultBiplot y la metodología de minería de datos. Ámbito Pacientes adultos politraumatizados ingresados en el Complejo Hospitalario de Salamanca entre los años 2006 y 2011. Variables de interés principales Variables demográficas, clínicas, terapéuticas y analíticas desde el lugar del accidente hasta el ingreso en la UCI. Variables evolutivas durante el ingreso en la UCI y hasta el alta hospitalaria. Resultados Se incluyó a 497 pacientes, con una mediana de edad 45,5 años. Predominio de varones (76,7%). La causa principal del traumatismo fueron los accidentes de tráfico (56,1%), precipitaciones (18,4%) y caídas (11%). Los factores con mayor asociación a un incremento del riesgo de mortalidad (p < 0,05) fueron la edad > 65 años (OR 3,15), el traumatismo craneoencefálico (OR 3,1), las alteraciones pupilares (OR 113,88), el nivel de consciencia según la escala de Glasgow ≤ 8 (OR 12,97) y las cifras de lactato > 4 mmol/L (OR 9,7). Conclusiones Los principales factores de riesgo identificados en relación con el pronóstico de los pacientes politraumatizados son los relacionados con la presencia de traumatismo craneoencefálico. Mediante la utilización de distintas técnicas estadísticas menos conocidas como la minería de datos o el MultBiplot también se destaca la importancia de otros factores como el lactato. Los registros de traumatismos ayudan a conocer la asistencia sanitaria realizada para poder establecer medidas de mejora.

An Architectural Multi-Agent System for a Pavement Monitoring System with Pothole Recognition in UAV Images

Wed, 01 Jan 2020 00:00:00 GMT

[EN]In recent years, maintenance work on public transport routes has drastically decreased in many countries due to difficult economic situations. The various studies that have been conducted by groups of drivers and groups related to road safety concluded that accidents are increasing due to the poor conditions of road surfaces, even affecting the condition of vehicles through costly breakdowns. Currently, the processes of detecting any type of damage to a road are carried out manually or are based on the use of a road vehicle, which incurs a high labor cost. To solve this problem, many research centers are investigating image processing techniques to identify poor-condition road areas using deep learning algorithms. The main objective of this work is to design of a distributed platform that allows the detection of damage to transport routes using drones and to provide the results of the most important classifiers. A case study is presented using a multi-agent system based on PANGEA that coordinates the different parts of the architecture using techniques based on ubiquitous computing. The results obtained by means of the customization of the You Only Look Once (YOLO) v4 classifier are promising, reaching an accuracy of more than 95%. The images used have been published in a dataset for use by the scientific community.

Avances en Informática y Automática. Decimoctavo Workshop

Tue, 10 Mar 2026 00:00:00 GMT

[ES]El Máster Universitario en Sistemas Inteligentes de la Universidad de Salamanca tiene como misión fundamental introducir a sus estudiantes en el rigor de la investigación científica. En este contexto, el congreso organizado por el Departamento de Informática y Automática se consolida como el escenario idóneo para que los alumnos presenten los resultados de sus Trabajos de Fin de Máster (TFM), sometiéndolos al análisis y debate propios de la comunidad académica. La decimoctava edición del workshop “Avances en Informática y Automática”, celebrada durante el curso 2024 - 2025, ha destacado por su marcado carácter multidisciplinar. En esta ocasión, las investigaciones presentadas abarcan un espectro tecnológico de vanguardia, incluyendo: - Procesamiento de Lenguaje Natural y LLMs: Desde la clasificación de textos en sectores como la hostelería hasta el ajuste fino de mo delos para detectar visualizaciones engañosas. - Machine Learning y Predicción: Aplicaciones prácticas en la previsión de demanda para retail, optimización energética en sistemas intralogísticos y control de alineamiento en sistemas láser de alta potencia. - Visión Artificial y Ciberseguridad: Comparativas de modelos para la detección de violencia y el desarrollo de sistemas IDS/IPS mediante deep learning. - Gestión de Datos y Sociedad: Análisis de espacios de datos en el sector Agrotech, desambiguación de autores en bases bibliográficas y herramientas de IA para facilitar el preprocesamiento de datos a usuarios no expertos. Bajo la supervisión de investigadores de prestigio de la Universidad de Salamanca, este encuentro no solo valida la calidad técnica de los trabajos, sino que sirve de puente hacia la realización de futuras tesis doctorales. Los objetivos principales del evento se mantienen firmes: - Exposición: Brindar a los estudiantes su primera experiencia formal en la difusión de resultados de investigación. - Intercambio: Crear un foro de discusión donde converjan ideas de compañeros, docentes y expertos. - Retroalimentación: Facilitar críticas constructivas que orienten las futuras líneas de investigación de los egresados. - Colaboración: Fortalecer el espíritu de trabajo conjunto y la sinergia entre diferentes áreas de conocimiento.

Inteligencia artificial fiable para la detección de violencia en vídeo

Wed, 01 Jan 2025 00:00:00 GMT

[ES] Las agresiones físicas son un problema grave y generalizado, como lo demuestra el hecho de que más de una cuarta parte (27%) de las mujeres de entre 15 y 49 años a nivel global declaran haber sido sometidas a algún tipo de violencia física y/o sexual por parte de su pareja íntima. La Inteligencia Artificial y específicamente las técnicas de Visión Artificial, ofrecen una solución eficaz para detectar la violencia en tiempo real, reduciendo la necesidad de supervisión humana constante. La Inteligencia Artificial, y en particular las técnicas de Visión Artificial, pueden contribuir a identificar episodios de violencia en tiempo real en lugares previamente delimitados, respetando los marcos éticos y legales establecidos. Sin embargo, el aumento del uso de la inteligencia artificial ha generado preocupación sobre la fiabilidad de los algoritmos, lo que ha llevado a la creación de informes destinados a establecer estándares y guías, con organizaciones como la Comisión Europea liderando estos esfuerzos. En este respecto, existen múltiples propuestas de algoritmos para la detección de violencia, donde la combinación de arquitecturas más comúnmente empleada es la de Redes Neuronales Convolucionales (CNN) y Redes de Memoria a Corto y Largo Plazo (LSTM), la cual obtiene excelentes resultados, si bien todavía persisten desafíos; sin embargo, hasta donde se conoce, ningún trabajo en el estado del arte ha abordado la detección de violencia mediante el uso de inteligencia artificial explicable, lo que limita la comprensión y confianza en los resultados obtenidos. Por ello, el objetivo principal de esta Tesis Doctoral es investigar, diseñar, desarrollar y validar algoritmos basados en técnicas de inteligencia artificial fiable orientadas en la detección de violencia en vídeo, con foco en arquitecturas basadas en la combinación de CNN junto con capas LSTM. En base a ello, en este trabajo se ha llevado a cabo un análisis y categorización de todos los procesos que involucran la detección de violencia en vídeo. Posteriormente se han investigado, diseñado, desarrollado y validado tres arquitecturas que utilizan la arquitectura VGG-19 preentrenada, una red neuronal convolucional conocida por su capacidad para extraer características visuales, combinadas con: características manuales, capas LSTM y capas Bi-LSTM. Por último, a partir de estas arquitecturas se han implementado técnicas de inteligencia artificial explicable como GradCAM y se ha creado un algoritmo que cuantifica el nivel de importancia para la detección de violencia por parte de las capas LSTM y Bi-LSTM. Los resultados obtenidos demuestran que el uso de capas Bi-LSTM supera al rendimiento obtenido por capas LSTM, si bien esta mejora no supera el 4% de exactitud. No se han encontrado valores o combinaciones de hiperparámetros para las arquitecturas que utilizan capas LSTM y Bi-LSTM que mejoren de una forma estadísticamente significativa la accuracy obtenida. Las arquitecturas desarrolladas han obtenido buenos reusltados como, por ejemplo, la combinación de VGG-19 preentrenada con capas Bi-LSTM, que obtiene un 97% de exactitud utilizando el dataset Hockey Fights. Por último, se ha conseguido hacer más explicable el proceso de detección con las técnicas implementadas.; [EN] Physical aggressions constitute a serious and widespread issue in society. Studies indicate that in 2015, at least half of the children in Asia, Africa, and North America experienced violence. Although solutions have been explored for medium and long-term interventions, real-time violence detection through artificial intelligence offers a direct and efficient solution that can save lives and reduce the need for constant human supervision. On the other hand, the increasing use of artificial intelligence has raised concerns about the development of reliable algorithms, leading to the creation of reports to define and standardize these terms. Major organizations such as the European Comission are leading this effort. There are multiple algorithm proposals for violence detection, with the most commonly employed combination being Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks, which yield excellent results. However, there are still issues to address, such as the actual impact of using LSTM layers instead of just CNN, how much violence detection improves with CNN combined with Bi-LSTM layers instead of LSTM layers, or if certain values and combinations of hyperparameters yield better results. Lastly, the use of reliable artificial intelligence remains very limited. Based on this, this work has developed a systematic literature review with the analysis and categorization of: 21 challenges associated with violence detection, 28 public datasets on violence v´ıdeos, and 13 evaluation metric methods; among others. Three architectures have been developed using pre-trained VGG-19 combined with: manual features, LSTM layers, and Bi-LSTM layers. It is evident that the use of Bi-LSTM layers outperforms the performance obtained by LSTM layers, although this improvement does not exceed 3% accuracy. No values or combinations of hyperparameters that significantly improve the obtained accuracy have been found statistically. The developed architectures have achieved good results, such as the combination of pre-trained VGG-19 with Bi-LSTM layers, which achieves 97% accuracy using the Hockey Fights dataset and 90% using the Violent Flow dataset. Lastly, the use of explainable artificial intelligence techniques on the proposed architectures, where YoloV8 and Frame Difference are used for the extraction of characteristic frames, GradCAM to highlight the areas VGG-19 focuses on for each convolutional layer, and a proprietary algorithm quantifies the level of importance for violence detection by LSTM and Bi-LSTM layers in violence detection.

Intelligent sensors in assistive systems for deaf people: a comprehensive review

Mon, 01 Jan 2024 00:00:00 GMT

[EN]This research aims to conduct a systematic literature review (SLR) on intelligent sensors and the Internet of Things (IoT) in assistive devices for the deaf and hard of hearing. This study analyzes the current state and promise of intelligent sensors in improving the daily lives of those with hearing impairments, addressing the critical need for improved communication and environmental interaction. We investigate the functionality, integration, and use of sensor technologies in assistive devices, assessing their impact on autonomy and quality of life. The key findings show that many sensor-based applications, including vibration detection, ambient sound recognition, and signal processing, lead to more effective and intuitive user experiences. The study emphasizes the importance of energy efficiency, cost-effectiveness, and user-centric design in developing accessible and sustainable assistive solutions. Moreover, it discusses the challenges and future directions in scaling these technologies for widespread adoption, considering the varying needs and preferences of the end-users. Finally, the study advocates for continual innovation and interdisciplinary collaboration in advancing assistive technologies. It highlights the importance of IoT and intelligent sensors in fostering a more inclusive and empowered environment for the deaf and hard-of-hearing people. This review covers studies published between 2011 and 2024, highlighting advances in sensor technologies for assistive systems in this timeframe.

JVM optimization: An empirical analysis of JVM configurations for enhanced web application performance

Mon, 01 Jan 2024 00:00:00 GMT

[EN]This research presents software for empirically analyzing Java Virtual Machine (JVM) parameter configurations to enhance web application performance. Using tools like JMeter and cAdvisor in a controlled hardware environment, it collects and analyzes performance metrics. Tailored JVM settings for high request loads improved CPU efficiency by 20% and reduced memory usage by 15% compared to standard configurations. For I/O intensive operations with large files, optimized JVM configurations decreased response times by 30% and CPU usage by 25%. These findings highlight the impact of tailored JVM settings on application responsiveness and resource management, providing valuable guidance for developers and engineers.

An analysis of the use of augmented reality and virtual reality as educational resources

Sun, 01 Jan 2023 00:00:00 GMT

[EN]In recent years, the utilization of augmented reality (AR) and virtual reality (VR) has emerged as a transformative approach in education, revolutionizing traditional teaching methods. This study seeks to explore the efficacy of AR and VR as pedagogical resources for enhancing the teaching of the solar system. The research process involved the development of an application comprising two modules, AR and VR, which were evaluated to assess their impact on the teaching process. Furthermore, a comparative study was conducted to evaluate the immersiveness, interactivity, and ease of use offered by these technologies. The findings demonstrate that both AR and VR demonstrate promise in supporting the teaching process, with the VR module garnering particularly positive evaluations. However, it is crucial to acknowledge existing barriers in underprivileged communities, where public schools face limited investments in technology infrastructure. These limitations hinder the widespread implementation of such immersive experiences and their potential to foster new knowledge acquisition.

Edge Face Recognition System Based on One-Shot Augmented Learning

Sat, 01 Jan 2022 00:00:00 GMT

[EN]There is growing concern among users of computer systems about how their data is handled. In this sense, IT (Information Technology) professionals are not unaware of this problem and are looking for solutions to meet the requirements and concerns of their users. During the last few years, various techniques and technologies have emerged that allow us to answer to the problem posed by users. Technologies such as edge computing and techniques such as one-shot learning and data augmentation enable progress in this regard. Thus, in this article, we propose the creation of a system that makes use of these techniques and technologies to solve the problem of face recognition and form a low-cost security system. The results obtained show that the combination of these techniques is effective in most of the face detection algorithms and allows an effective solution to the problem raised.

Active Actions in the Extraction of Urban Objects for Information Quality and Knowledge Recommendation with Machine Learning

Fri, 23 Dec 2022 00:00:00 GMT

[EN]Due to the increasing urban development, it has become important for municipalities to permanently understand land use and ecological processes, and make cities smart and sustainable by implementing technological tools for land monitoring. An important problem is the absence of technologies that certify the quality of information for the creation of strategies. In this context, expressive volumes of data are used, requiring great effort to understand their structures, and then access information with the desired quality. This study are designed to provide an initial response to the need for mapping zones in the city of Itajaí (SC), Brazil. The solution proposes to aid object recognition employing object-based classifiers OneR, NaiveBayes, J48, IBk, and Hoeffding Tree algorithms used together with GeoDMA, and a first approach in the use of Region-based Convolutional Neural Network (R-CNN) and the YOLO algorithm. All this is to characterize vegetation zones, exposed soil zones, asphalt, and buildings within an urban and rural area. Through the implemented model for active identification of geospatial objects with similarity levels, it was possible to apply the data crossover after detecting the best classifier with accuracy (85%) and the kappa agreement coefficient (76%). The case study presents the dynamics of urban and rural expansion, where expressive volumes of data are obtained and submitted to different methods of cataloging and preparation to subsidize rapid control actions. Finally, the research describes a practical and systematic approach, evaluating the extraction of information to the recommendation of knowledge with greater scientific relevance. Allowing the methods presented to apply the calibration of values for each object, to achieve results with greater accuracy, which is proposed to help improve conservation and management decisions related to the zones within the city, leaving as a legacy the construction of a minimum technological infrastructure to support the decision.

Time-Constrained Ontology Evolution for Predictive Maintenance

Wed, 01 Jan 2025 00:00:00 GMT

[EN] With the introduction of the Internet of Things, maintenance practices have been moving from reactive to proactive and predictive approaches. The identification of faults often relies on the analysis of real-time data provided by streams and unstructured sources. Ontologies have been applied to the maintenance field, adding a semantic layer to the data that facilitates interoperability and semantic data mining processes. In such a time-sensitive domain, it is important that ontologies go beyond static representations of the domain and allow not only for the incorporation of time related knowledge, but must also be able to adapt to new knowledge and evolve. Evolving an ontology involves re-learning, re-enriching and re-validating knowledge in the face of changes to the domain, and techniques applied for them can be adapted to ontology evolution. This thesis aims to contribute to these fields by using streams of ontology individuals as the trigger for ontology evolution processes – facing challenges tied to the incomplete and transient nature of these data. As such, this thesis introduces an architecture for time-constrained ontology evolution called TICO, or Time Constrained instance-guided Ontology evolution. New versions of ontology classes and properties are reified through a 4D-Fluents approach, thus allowing reasoning over old data and accessing older conceptualizations of the domain. For the identification of property axioms, the possibilistic approach to axiom scoring was adapted to a scenario in which it is not always possible to query all individuals at once. Results show the effectiveness of the approach in accepting/rejecting axioms for the ontology’s properties. To identify patterns in data that could trigger the creation of new classes and enrich existing ones, a Formal Concept Analysis-based approach is employed. Using two different concept lattices that are updated with each individual, it is possible to identify a set of axioms to add to the ontology and uncover implicit relationships between old and new classes.; [ES] Con la introducción del IoT, las prácticas de mantenimiento han ido pasando de orientaciones reactivas a proactivas y predictivas. La identificación de fallas a menudo se basa en el análisis de datos en tiempo real proporcionados por flujos y fuentes no estructuradas. Las ontologías se han aplicado al campo del mantenimiento, añadiendo una capa semántica a los datos que facilita la interoperabilidad y los procesos de minería semántica de datos. En un ámbito tan sensible al tiempo, es importante que las ontologías ultrapasen las representaciones estáticas del dominio y permitan no sólo incorporar conocimientos relacionados con el tiempo, sino que también deben ser capaces de adaptarse y evolucionar. Evolucionar una ontología implica reaprender, re-enriquecer y re-validar el conocimiento y las técnicas aplicadas para ellas pueden adaptarse a la evolución de ontologías. Esta tesis pretende contribuir a estos campos utilizando flujos de individuos RDF como desencadenante de procesos de evolución de ontologías, enfrentándose a retos ligados a la naturaleza incompleta y transitoria de estos datos. Como tal, esta tesis introduce una arquitectura para la evolución de ontologías limitada en el tiempo llamada TICO (Time Constrained instance-guided Ontology evolution). Las nuevas versiones de las clases y propiedades de la ontología se reifican mediante 4D-Fluents, lo que permite razonar sobre datos antiguos y acceder a conceptualizaciones anteriores del dominio. Para la identificación de axiomas de propiedades, se adaptó el enfoque posibilista de cualificación de axiomas a un escenario en el que no siempre es posible obtener la descripción completa del conjunto de datos. Los resultados muestran la eficacia de la solución en aceptar/rechazar axiomas para las propiedades de la ontología. Para identificar patrones en los datos, crear nuevas clases y enriquecer las existentes, se emplea un enfoque basado en el Análisis Conceptual Formal. Utilizando dos redes de conceptos diferentes, es posible identificar un conjunto de axiomas para añadir a la ontología y descubrir relaciones implícitas entre clases antiguas y nuevas.

Efficient Suboptimal Detectors for Maritime Surface Surveillance High-Resolution Radar

Fri, 24 Aug 2012 00:00:00 GMT

[EN] This paper presents some efficient suboptimal detectors, based on statistical descriptors, which take advantage of the high-resolution characteristics of the high-resolution radars (HRR). Which are one of the first stages of the sensor-based localization and tracking technologies. The detection performance has been studied under noise and sea clutter conditions, with non-coherent data from both real and synthetic extended targets. We have also made an adaptation of the classical moving window detection technique for the high-resolution radars, making use of it as a reference technique to evaluate the results obtained with the detection techniques that we present. The experimental results were obtained with the ARIES radar, a maritime surface surveillance LFM-CW HRR operating in X-band.