http://repositorio.ufc.br/handle/riufc/478 2026-04-07T13:17:32Z http://repositorio.ufc.br/handle/riufc/85536 Título: A study on the angle-of-arrival estimation problem from three complementary perspectives: challenges, particularities, and applications Autor(es): Santos, Michel Gonzaga dos Abstract: This PhD dissertation condenses the outcomes of the studies developed over the course of this doctoral program. It begins with an overview chapter that provides the theoretical background required to support the topics addressed throughout the thesis. The remaining chapters are connected through the central theme of this thesis: the angle of arrival (AoA) estimation problem in wireless systems, which is explored under three distinct yet complementary perspectives. The main motivation for writing this document is to support the argument that a clear understanding of the AoA estimation problem is a key enabler for advancing the technologies envisioned for 6th generation (6G) and beyond, by bridging the communication and radar sensing domains. Accordingly, gaining insight into the nuances of this problem is of great importance, as it provides the foundation for fully exploiting the potential of both services. First, the problem is examined from a purely communication perspective, in which a novel AoA estimation approach for analog beam refinement is proposed, exploiting the polarization domain in high-frequency multiple-input multiple-output (MIMO) communication networks. Subsequently, the joint AoA and Doppler frequency estimation problem is investigated from a sensing perspective. By considering bistatic and multistatic sensing networks, the joint estimation of these two parameters enables target localization capabilities within the system. In this context, the Cramér-Rao lower bound (CRLB) is employed to assess the performance of the individual estimators under different scenario setups. Finally, the AoA and angle of departure (AoD) estimation problem is examined in a bistatic integrated communication and sensing (ISAC) network. Assuming downlink transmission, it is argued that the quality-of-service (QoS) of both sensing and communication systems can be improved through a proper allocation of space–time resources. To this end, an optimization problem is formulated to jointly enhance the time-averaged communication sum-rate—assuming a given number of users (UEs)—and the CRLB associated with the AoA and AoD of the passive target of interest. Additionally, the main conclusions derived from these chapters, along with directions for future research, are summarized in the Conclusion chapter. Tipo: Tese 2025-01-01T00:00:00Z http://repositorio.ufc.br/handle/riufc/85335 Título: SenseAir: arquitetura de monitoramento de qualidade do ar com sensores de baixo custo e visão computacional para tráfego veicular Autor(es): Monteiro, Nícolas de Carvalho Abstract: The scarcity and regional concentration of reference air quality monitoring stations in Brazil limit reliable diagnostics and the design of evidence-based public policies, especially in undermonitored regions such as the Northeast. As a complementary alternative to official networks, this dissertation investigates a low-cost architecture composed of an embedded sensor node (SenseAir), a real-time mobile application, a computer vision model for vehicle traffic quantification, and an exploratory calibration module for PM2,5 anchored in data from a higher-performance portable instrument. The solution integrates continuous data acquisition, secure transmission and cloud storage, and visualization of the Air Quality Index (AQI) derived from PM2,5 with geolocation, targeting indicative monitoring in urban scenarios. Calibration is discussed in light of U.S. EPA guidelines for PM2,5 sensors, using metrics such as coefficient of determination (R2), bias (regression slope and intercept), and root mean square error (RMSE) to evaluate the correction applied to SenseAir. In parallel, a YOLO-based model embedded in the Raspberry Pi estimates the flow and composition of the vehicle fleet, enabling analysis of the role of traffic, together with other external factors, in the local degradation of air quality. The results discuss cost, performance, and limitations of the proposed architecture, as well as its potential use in dense indicative monitoring networks and in research and decision-support applications for public management. Tipo: Dissertação 2026-01-01T00:00:00Z http://repositorio.ufc.br/handle/riufc/85317 Título: Entropia diferencial multiescala e agregação de descritores na classificação de folhas de plantas Autor(es): Pinheiro, Raphael Gomes Abstract: Shape analysis and recognition are fundamental in the design of computer vision-based systems. The greatest challenge is to develop robust methods capable of extracting significant features from shapes in order to represent them effectively. In this context, multiscale descriptors represent a versatile and efficient alternative for shape characterization. This work presents a methodology for classifying plant leaves based on handcrafted features derived from the multiscale entropy of curvature and texture, as well as deep features obtained from convolutional neural networks (CNNs). We propose three object descriptors based on the multiscale entropy of curvature. These descriptors rely on the differential entropy of the probability distributions of multiscale curvatures to create a coarse-to-fine representation of the shape contour. Furthermore, we present a descriptor that aggregates the multiscale entropy of curvature, bending energy of curvature, and texture features to enhance the extraction of object signatures and subtle texture details in leaf images. The texture descriptor combines statistics from the local binary pattern and gray-level co-occurrence matrix. We compare our handcrafted descriptors with deep features extracted from various CNNs in a multiclass classification framework using the random forest classifier, replacing the fully connected layer of the CNNs with this classifier. Experiments were conducted on four public leaf datasets: Plantscan, MED117, Flavia, and Swedish. The results, evaluated using F1-score and accuracy metrics exceeding 99.50%, validate the aggregation strategy and demonstrate that it is competitive and robust. The findings also confirm that the proposed strategy outperformed four different sets of deep features according to both F1-score and accuracy metrics. Qualitative analysis through multidimensional data visualization further demonstrates that combining different shape features and texture details improves the description of leaf images, providing better intraclass compactness and interclass separation across the datasets. Tipo: Tese 2025-01-01T00:00:00Z http://repositorio.ufc.br/handle/riufc/85315 Título: VLBODY — Vision-learned body composition estimator: uma abordagem profunda para análise visual da composição corporal Autor(es): Ivo, Roberto Fernandes Abstract: Accurate assessment of body composition is essential for monitoring nutritional status, physical performance, and metabolic risk in individuals. Traditional methods, such as dual-energy X-ray absorptiometry (DXA), although highly accurate, present limitations related to cost, accessibility, and large-scale applicability. In this context, this thesis proposes VLBODY — Vision-Learned Body Composition Estimator, a deep visual learning methodology designed for the noninvasive estimation of body fat percentage from two-dimensional images. The proposed method is organized as a sequential pipeline comprising three main stages: (A) detection, responsible for the automatic identification of the region of interest (ROI) corresponding to the human body in the image; (B) segmentation, aimed at the precise isolation of the body silhouette and removal of background noise; and (C) body fat estimation, which relies on the visual encoding component of Vision–Language Model (VLM) architectures to extract morphological representations from the ROI and, through a regression process, convert them into continuous body fat percentage values. This unified architecture integrates the entire process, from image input to final prediction, ensuring coherence between feature extraction and quantitative inference. Four architectural variants (VLBODY-S, VLBODY-M, VLBODY-L, and VLBODY-X) were implemented, with different depths and latent dimensions, to investigate the impact of representational capacity on predictive performance. The method was validated on the BCDB23 dataset, composed of 1,044 participants whose standardized images were compared against DXA reference measurements. The VLBODY-S and VLBODY-X variants achieved the best results, reaching coefficients of determination (R²) above 0.8 and mean absolute errors (MAE) below 3 percentage points, outperforming traditional anthropometric-based methods and approaching state-of-the-art techniques. In addition to quantitative performance, VLBODY incorporates an interpretability mechanism based on spatial relevance maps, highlighting the body regions that most influence predictions. These results indicate that VLBODY represents a promising alternative for noninvasive body composition assessment, combining accuracy, interpretability, and potential applicability in both clinical and population-based contexts. Tipo: Tese 2025-01-01T00:00:00Z