1,725,787 research outputs found
Multiview Video Coding Accelerated on Multicore Architectures
No full textThis thesis deals with the design and implementation of extremely parallel fast motion / disparity estimation algorithm for multicore architectures. Currently, H.264/AVC is the most widely used commercial video compression standard and is based on single view. Recently, Multi-view Video Coding (MVC) has also been standardized as an extension to H.264/AVC for supporting 3D and Free Viewpoint video. As MVC is an extension to H.264/AVC, so it achieves compression not only by exploiting temporal and spatial prediction but also exploits inter-view redundancies using motion estimation tool. In H.264/AVC, motion estimation is the most important tool employed by the video encoder to mitigate temporal redundancies but it is also the most time consuming. Consequently, in MVC, the time consumed for efficient encoding is even higher as the encoder has to perform temporal as well as inter view predictions. This thesis proposes a parallel low-complexity rate-distortion optimized motion/disparity estimation algorithm that can be implemented on multicore architectures such as Graphical Processing Unit (GPU). Recently, GPU has emerged as a commercially viable multicore platform for accel-
erating computationally extensive applications and has also been applied for improving video encoder performance. Generally, the bit rate cost during motion vector calculation is ignored while implementing parallel motion estimation algorithms on GPU, due to the unavailability of the spatially predicted motion vectors, which leads to rate-distortion performance degradation. The proposed approach is able to perform the complex prediction task by means of an efficient distribution of all the computations over the GPU by mitigating the spatial dependencies. The experimental results show that the proposed scheme achieves significant speedup and has comparable rate-distortion performance with respect
to sequential fast motion estimation algorithm. The proposed algorithm is also used for exploiting inter-view prediction in MVC and is implemented on the GPU exploiting view and block level parallelism simultaneously. The results for MVC suggest a significant speedup with negligible loss in coding efficiency
Improving Quality of Results for High-Level Synthesis based FPGA designs
No full textL'abstract è presente nell'allegato / the abstract is in the attachmen
Hybrid Rigid–Soft Industrial Gripper: Actuation, Design Enhancement, Multi-Modal Sensorization, and Real-Time Coordinated Control for Automotive Assembly
No full textThe growing complexity of industrial automation and the shift toward human–robot collaborative manufacturing demand robotic grippers that combine mechanical precision, adaptive compliance, and intelligent sensing. This doctoral thesis addresses these needs through comprehensive research on actuation mechanisms, kinematic analysis, design enhancement, advanced sensorization, and industrial implementation of a versatile universal gripper system capable of handling components from delicate items to complex rigid parts. In parallel, it contributes to industrial robot control within the EU Horizon SESTOSENSO project.
The work begins with a detailed kinematic analysis of a novel three-finger gripper architecture. Each rigid mechanical finger integrates a Chebyshev–parallelogram linkage mechanism with a thermoplastic polyurethane (TPU) contact interface. The mechanism produces near-linear trajectories with a deviation of ±0.033 mm and a mechanical advantage of 6.06:1. Independent actuation is achieved using JVL stepper motors with embedded programmable logic controllers (PLCs), communicating via Modbus remote terminal unit (RTU). The control architecture supports torque-based and velocity-based stall detection, both operating in real time with configurable thresholds. These strategies enable reliable grasping without dedicated force sensors by leveraging internal motor feedback parameters.
Gripper enhancement is guided by a quantitative deflection coefficient to assess finger wrapping and by quasi-static force–displacement testing. The original four-bar parallelogram was redesigned into a six-bar linkage with compliant pads. This eliminates link interference limitations while preserving the essential kinematics, resulting in adaptive grasping capability. Pull-out tests demonstrated improved force profiles for complex automotive parts and reliable manipulation of objects from 100 g to 7.5 kg.
Vision-based sensorization was achieved through an embedded Raspberry Pi Camera V3.
The integration of vision-based sensing within the additively manufactured soft finger structure establishes the feasibility of achieving multiple sensing modalities with a single compact embedded system while retaining the characteristic properties of the fingers. The proposed system successfully estimates normal interaction forces, measures internal deformation (Z-displacement), classifies the position of the applied force, and detects slip events with the complete sensing pipeline processed on an embedded platform while avoiding complex signal disambiguation challenges and occlusion issues. Complementing this, a fully flexible resistive sensor was fabricated via fused deposition modeling (FDM) printing and embedded in the finger for contact and bending detection. A novel light-angle sensor array was also developed using a custom four-layer rigid-flex printed circuit board (PCB), where prototype sensors successfully demonstrate distributed tactile sensing capabilities.
The universal gripper and sensing systems were validated on a COMAU six-degrees-of-freedom (6-DOF) industrial robot in diverse grasping trials, confirming adaptability, robustness, and sensing reliability. Separately, within the SESTOSENSO project, real-time control strategies were developed for coordinating a KUKA KR150 robot with a UR10 cobot via robot sensor interface (RSI) and robot operating system (ROS) in an automotive roof assembly task. This work addressed control architecture, real-time trajectory correction, and safe human–robot collaboration in confined, visually occluded environments.
This thesis advances the state of the art in hybrid gripper systems by integrating rigid precision, soft adaptability, and intelligent sensing with industrially validated control strategies. The outcomes directly support Industry 4.0/5.0 objectives, enabling flexible, high-performance automation adaptable to diverse manufacturing requirements
Advancing Plastic Recycling : A Review on the Synthesis and Applications of Hierarchical Zeolites in Waste Plastic Hydrocracking
Full text linkAcknowledgments: This work was funded by The LEVERHULME TRUST (Grant DS-2017-073). Muhammad Usman Azam, a Leverhulme Trust Doctoral Scholar, was part of the 15 PhD scholarships of the “Leverhulme Centre for Doctoral Training in Sustainable Production of Chemicals and Materials” at the University of Aberdeen (Scotland, United Kingdom).Peer reviewe
Deep Learning Approaches Targeting Radiological Images
No full textArtificial Intelligence (AI) algorithms have remarkably improved their performance in the recent years
in various domains, thanks to the introduction of deep learning approaches. Indeed they have shown a tremendous potential when solving tasks involving image analysisThe problem of deep learning is its requirement for huge datasets, nonetheless, DL approaches have proved to be helpful in the domain of medical imaging as well. Automated segmentation and classification in different biomedical tasks have proven to be faster and more cost effective.
In this thesis we study deep learning approaches used for segmentation and classification of different radiological images mainly CT Scans, MRI Scans and CXR images. In particular, we explored some issues like the multi-modality, and the small dataset problem
We first discuss about how the small datasets can be exploited to improve the performance of the deep model in the proposed architectures and then in the next work we train the model with multi modal data consisting of both CT and MRI images together and consider the corresponding opposite modality of CT and MRI as missing data problem. We use Cycle-GAN to generate the synthetic data for the missing data and further train the model with original and synthetic data together.
Then we focus on the classification of COVID exploiting the multi-modality data available. We proposed an architecture that is capable of handling multi modal data and extract feature representation from available modalities before concatenation and further use them for final classification. Then we exploit joint learning to train a small dataset from scratch.
Finally, this thesis concludes with open questions that may benefit from future work. This thesis demonstrate the potential role of CNNs to address the tasks of segmentation and classification
Evaluation of the effect of climate variability on the hydro-glaciological regime in the Upper Indus Basin
Full text linkIl ritiro dei ghiacciai osservato a scala planetaria e gli effetti del riscaldamento globale che accelererà la fusione dei ghiacciai e delle nevi stagionali andranno ad alterare alterare i regimi idrologici che interessano non solo i bacini montani, ma anche le aree a valle. Gran parte dell'approvvigionamento idrico del Pakistan è generato dalla fusione nivale e glaciale nelle regioni montuose del Karakoram. Tuttavia, il cambiamento climatico rappresenta un rischio elevato per questi bacini idrografici. Pertanto, quantificare questi cambiamenti prevedibili è una sfida importante per la gestione e la pianificazione delle risorse idriche. L'obiettivo di questa tesi è valutare l'effetto della variabilità climatica sul regime idrologico e glaciologico nell’Alto Bacino dell’Indo (nel seguito Upper Indus Basin-UIB).
Il primo capitolo di questa tesi di dottorato presenta la climatologia delle precipitazioni ad alta risoluzione nell’UIB basata sul metodo delle anomalie (1995-2017) e sviluppato utilizzando quattro set di dati su griglia (APHRODITE, CHIRPS, PERSIANN-CDR e ERA5) su scala stagionale e annuale. I risultati indicano le migliori prestazioni della precipitazione stimata con il dataset CHIRPS con correzione del bias seguito dalle rianalisi del modello ERA5; infatti il set di dati di precipitazione CHIRPS con correzione del bias ha ottenuto risultati migliori nella simulazione delle precipitazioni con RMSE, MAE, MAPE [%] e BIAS più piccoli seguiti da ERA5.
Sulla base dei risultati del Capitolo 1, nel Capitolo 2 si utilizza il modello distribuito di bilancio energetico Physically Based Distributed Snow Land and Ice Model (PDSLIM), già testato nelle Alpi da Ranzi e Rosso (1991), Ranzi et al. (2010) e Grossi et al. (2013) per il bacino del Naltar situato nel bacino del fiume Hunza, in Pakistan, per simulare i regimi idro-glaciologici attuali e futuri. I risultati hanno mostrato prestazioni molto soddisfacenti del modello verificato rispetto all'area di copertura nevosa (Snow Cover Area) stimata dalle immagini del sensore satellitare LANDSAT-TM e Terra/Aqua-MODIS per tutti gli anni simulati con coefficiente medio di determinazione R2 = 0,96 e Nash-Sutcliffe Efficiency NSE= 0,95. Le simulazioni di deflusso hanno rivelato un buon accordo con la portata giornaliera osservata ottenuta con NSE e Kling-Gupta Efficiency (KGE) medi di 0,90 e 0,89. L'analisi della composizione del deflusso ha rivelato che la componente sotterranea, con risposta lenta, è la componente principale, seguita dal deflusso del ghiacciaio e dal deflusso superficiale.
Nel Capitolo 3 si illustra l’impiego del modello PDSLIM calibrato per esaminare le proiezioni future dei regimi glaciologico-idrologici per i due periodi temporali (2040-2059) e (2080-2099) negli scenari RCP 2.6, RCP 4.5 e RCP 8.5. Le simulazioni previste del bilancio di massa ed energetico indicano che la progressione della fusione della neve e del ghiaccio aumenterà costantemente in entrambi i periodi di tempo futuri con un'anticipazione dei tempi della massima fusione nivale. Dalle stime del bilancio di massa attuale (-737 mm anno-1) e previsto (-887 mm anno-1 per lo scenario 2050_4.5; -2018 mm anno-1 per il 2050_8.5 e -1154 mm anno-1 entro il 2090_4.5 ; -2597 mm anno-1 per 2090_8.5) e dalle immagini satellitari MODIS e LANDSAT sembra che anche nel bacino del Naltar i ghiacciai stiano per ritirarsi rapidamente indicando un'eccezione alla cosiddetta 'anomalia del Karakoram', una congettura di un rallentamento del ritiro dei ghiacciai nella regione a causa dell'accumulo nivale indotto dalle precipitazioni ad alta quota. Nel complesso il modello PDSLIM mostra prestazioni molto buonei nel simulare le dinamiche glacio-idrologiche attuali e, probabilmente, anche quelle future e pone una solida base per il potenziale utilizzo dell'approccio del bilancio energetico distribuito nei bacini glaciali del Karakorum e del’Hymalaia.Projections of glaciers’ retreat and earlier snowmelt driven by global warming could alter the hydro-glaciological regimes affecting not only the upstream watershed but also downstream areas. A large portion of Pakistan's water supply is generated by the melting of snow and ice in the mountainous regions of the Karakoram. However, climate change poses a high risk to these watersheds. Thus, quantifying these changes at the right time is an important challenge for water resources planners. The objective of this dissertation is to assess the effect of climate variability on the hydro-glaciological regime in the Upper Indus Basin (UIB) in the present and future projected climate.
Chapter 1 compiles the high resolution precipitation climatology (1995-2017) of the UIB developed using the anomaly method and four gridded datasets (APHRODITE, CHIRPS, PERSIANN-CDR and ERA5) which are bias corrected with interpolated observations at seasonal and annual scale. The results indicate the better performance of bias corrected CHIRPS precipitation followed by ERA5; bias corrected CHIRPS precipitation datasets performed better in simulating precipitation with smaller RMSE, MAE, MAPE [%] and BIAS followed by ERA5. Precipitation and discharge revealed significant variability at the seasonal scale more than at annual scale. The rainfall and runoff relationship and annual runoff coefficients suggest the need of further investigation and monitoring about snow-glacier melt contribution in streamflow.
Based on Chapter 1 outcomes, Chapter 2 employs the Physically Based Distributed Snow Land and Ice Model (PDSLIM), already tested in the Alps by Ranzi and Rosso (1991), Ranzi et al. (2010) and Grossi et al. (2013); Ranzi and Rosso (1991) for the Naltar catchment situated in the Hunza river basin (Pakistan) to simulate current and future hydro-glaciological regimes. The results exhibited very satisfactory performances of the model verified against satellite-based snow cover area for all simulated years with average coefficient of determination R2 = 0.96 and Nash-Sutcliffe Efficiency NSE= 0.95. Runoff simulations revealed good agreement with observed daily discharge obtained with mean NSE and KGE of 0.90 and 0.89.
Chapter 3 employs the calibrated PDSLIM to examine future projections of glaciological-hydrological regimes for the two-time periods 2040-2059 and 2080-2099 under RCP 2.6, RCP 4.5 and RCP 8.5 scenarios. The projected simulations of the energy and mass balance indicate that snow and ice melt progression will consistently increase in both future time periods with an anticipation in the timing of the maximum snowmelt. Additionally, the rise in temperature is expected to have a substantial impact on peak hydrological regimes from one to two months earlier by 2090s over Naltar catchment. From the actual (-737 mm a-1) and projected mass balance estimates (-887 mm a-1 by 2050_4.5 scenario; -2018 mm a-1 for 2050_8.5 and -1154 mm a-1 by 2090_4.5; -2597 mm a-1 for 2090_8.5) and the MODIS and LANDSAT satellite images it appears that also in the Naltar catchment glaciers are going to retreat fast indicating an exception to the so-called ‘Karakoram anomaly’, a conjecture of a slower retreat of glaciers in the region because of accumulated precipitation at high altitudes . Overall, PDSLIM performs well for the current and, likely, future glacio-hydrological dynamics and sets a strong foundation for the potential usage of distributed energy balance approach in the glacierized catchments of High Mountain Asia (HMA) including Karakoram and Himalaya
Multimodal Segmentation of Medical Images with Heavily Missing Data
No full textAn important aim of research in medical imaging is the development of computer aided diagnosis (CAD) systems. A fundamental step in these systems is the image segmentation and convolutional neural networks (CNNs) are becoming the most commonly used approach to solve this task. However, despite their great power, in this domain CNNs are limited in their potential performance by the usually small amount of data [1]. Computed tomography (CT) and magnetic resonance imaging (MRI) scans are often used to examine the internal structure of human body and have their own unique properties and limitations. As a common practice, the investigations are usually done on a single modality, nonetheless, the simultaneous analysis of multiple modalities can significantly boost the segmentation accuracy. However, obtaining multiple imaging modalities for the same subject is very unlikely. In this paper we investigate the possibility of generating a multimodal CT-MRI representation for a segmentation task starting from a single modality, either CT or MRI. We considered this as a missing data problem, hence, we designed a pipeline where a CycleGAN was used to generate the missing modality. The synthetic modality was then paired with the real one to perform the required segmentation taking advantage of the multimodal representation and the augmented training dataset. To test the system we used two unrelated labeled datasets, one with CT data and the other one with MRI data. Results show that data enrichment with synthetic modalities improves the segmentation performance
Parallel Rate-Distortion Optimised Fast Motion Estimation Algorithm for H.264/AVC using GPU
No full textEnumeration and Identification of Active Users for Grant-Free NOMA Using Deep Neural Networks
Full text linkIn next-generation mobile radio systems, multiple access schemes will support a massive number of uncoordinated devices exhibiting sporadic traffic, transmitting short packets to a base station. Grant-free non-orthogonal multiple access (NOMA) has been introduced to provide services to a large number of devices and to reduce the communication overhead in massive machine-type communication (mMTC) scenarios. In grant-free communication, there is no coordination between the device and base station (BS) before the data transmission; therefore, the challenging task of active users detection (AUD) must be conducted at the BS. For NOMA with sparse spreading, we propose a deep neural network (DNN)-based approach for AUD called active users enumeration and identification (AUEI). It consists of two phases: firstly, a DNN is used to estimate the number of active users; then in the second phase, another DNN identifies them. To speed up the training process of the DNNs, we propose a multi-stage transfer learning technique. Our numerical results show a remarkable performance improvement of AUEI in comparison to previously proposed approaches
Hydrocracking of surgical face masks over Y zeolites : catalyst development, process design and life cycle assessment
Full text linkAcknowledgement This study was funded by The LEVERHULME TRUST (Grant DS-2017-073). Muhammad Usman Azam, a Leverhulme Trust Doctoral Scholar, is part of the 15 PhD scholarships of the “Leverhulme Centre for Doctoral Training in Sustainable Production of Chemicals and Materials” at the University of Aberdeen (Scotland, United Kingdom). Auguste Fernandes thanks Portuguese FCT for funding (CQE - UIDB/00100/2020 and UIDP/00100/2020; IMS-LA/P/0056/2020; contract hiring under DL57/2016 law).Peer reviewe
- …
