1,721,030 research outputs found
Experimental comparison of some scheduling disciplines fed by self-similar traffic
No full textSelf-similar traffic models have permitted a more realistic description of the network devices behavior. However, the derivation of analytical results turns out to be a very demanding task, also in the single-server case. For the work-conserving switching architectures the characterization of the quality of service (QoS) parameters is even more complicated due to the correlation among the queues, induced by the scheduling policies. In this paper we present a detailed study, based on simulations of some paradigmatic scheduling disciplines, performed with an aim to furnish some useful tools for the design of high-speed network devices
Reduced And Full-Scale Assessment Of Super-Resolution Of Sentinel-5P Radiance Images
No full textThe spatial resolution of TROPOMI, the sensor mounted on board of the satellite Sentinel-5P to monitor air quality, is much higher than its predecessors. Yet, the high variability of pollutants limits the use of the resulting maps in practical applications. Super-resolution approaches can improve the precision of estimates, but their use is heavily reliant on the ability to precisely tune the parameters of the algorithms. For this reason, the employment of a specific image acquisition model is essential for both learning-based and traditional methods. This contribution leverages real full-scale images for validation of a recently proposed model for the degradation introduced by the TROPOMI instrument, which is applied to both classical and learning-based techniques. The model's validity can be evaluated by analysing the quantitative data and visually inspecting the images that have been generated. This contribution proves that the degradation model is an essential basis for the development of novel approaches as well as for the application of all already available techniques
A Study of the Relationships Between the Real Scene Statistics and Those of the Backscattered Signal
No full textPopular mathematical description of natural landscapes rely
upon fractal geometry and a peculiar parameter of the model is the Hurst coefficient HX, which rules the correlation properties of the real scene.
The random process modelling the remotely collected data may preserve the fractal behavior of the original scene and its second-order statistics is then characterized by a Hurst number HY . However, HY = HX is only one of the possibilities arising from the mapping real scene → collected data. The relationships between the two Hurst parameters, hence between the second-order properties of the correspondent random processes, are investigated in the simplified scenario where the above mapping is a zeromemory nonlinearity. The obtained results improve and corroborate the work of [2]
Efficient Hyperspectral Super-Resolution of Sentinel-5P Data via Dynamic Multidirectional Cascade Fine-Tuning
No full textSentinel-5P is a valuable resource for academics and policymakers. The ability of the satellite's equipment to span the electromagnetic spectrum from ultraviolet (UV) to short-wave infrared (SWIR) frequencies is vital in determining the distribution of important gaseous pollutants on a global scale, a significant turning point for air quality monitoring. In technical terms, Sentinel-5P provides an excellent balance between spatial and spectral resolutions; however, physical limitations keep hindering the quality of its products. S5Net is the first deep-learning-based (DL-based) approach designed to super-resolve Sentinel-5P radiance images. Despite its simplicity, this neural network has showed excellent performance when applied to monochromatic images, particularly when compared to more complex deep neural networks. Yet, this groundbreaking study has a significant limitation: the computational inefficiency of the fine-tuning employed, which must be adequately extended to numerous channels. We hence propose a novel dynamic multidirectional cascade fine-tuning procedure, whose routine is fully governed by the correlation between consecutive spectral channels. Our study is accordingly successful in striking a remarkable balance between spectral coherence and spatial resolution improvement, as well as substantially optimizing computing efficiency. The code is available at https://github.com/alcarbone/S5P_SISR_Toolbox
Super-resolution techniques for Sentinel-5P products
No full textAir pollution is considered a very critical environmental risk to human health. The World Health Organization reports that it is responsible for almost 7 million deaths. As so, motivation is enough to decrease population exposure. However, several unsolved issues that require additional research remain. In particular, despite global monitoring development, coverage is insufficient to accurately describe the spatial variability for specific pollutants within different areas. The TROPOshperic Monitoring Instrument mounted on Sentinel-5P is one of the satellite instruments that retrieve atmospheric pollutants' concentration with a comparatively high spatial resolution, around 5 km. However, the spatial detail of the available products is often unsuitable for the purpose at hand. Also, physical constraints prevent enhancing the sensor's nominal spatial resolution further. So, there is no alternative way to collect high-resolution information than through processing algorithms. In this research, we investigated the problem of super-resolving Sentinel-5P products by employing traditional and deep learning-based approaches. While the former do not require a training phase because they rely on simple physical models, the latter can attain higher performance by reproducing highly complicated models. However, the lack of high-resolution reference data makes the needed training phase of network parameters extremely challenging. In this paper, we studied different approaches tailored to the imagery at hand and evaluated their accuracy with Sentinel-5P data. This study provides insights into the techniques and how they should be employed to monitor air quality accurately. The results of this work give significant information for the development of suitable super-resolution algorithms
Rao-Blackwellised Particle Filter for Battery State-Of-Charge and Parameters Estimation
No full textState-of-charge and parameters online estimation is one of the key features of battery management systems for hybrid-electric vehicles applications. Using model-based approaches, simultaneous sequential Bayesian estimation of battery state and parameters has been shown to be a very powerful tool for the tracking, even in the presence of non-perfectly known models. Monte Carlo implementations are very suited to strongly nonlinear and unreliable dynamics, such those of batteries. In this framework, current paper proposes the use of a Rao-Blackwellized Particle Filter (RBPF) for the joint estimation of battery state and parameters. The results are compared with the existing approaches, highlighting the appealing features of RBPF, both in terms of performances and robustness
Comparing Particle Filter and Extended Kalman Filter for Battery State-Of-Charge Estimation
No full textThe battery State-Of-Charge (SOC) and parameters estimation is one of the crucial points to be addressed in the development of innovative electric/hybrid electric vehicles. Extended Kalman Filter (EKF) and Particle Filters (PF) are two possible approaches to the problem. While EKF is attractive for its computational efficiency, it may not be accurate for the non-linearity and for the uncertainties involved in the battery modelling. PF is a promising alternative, even if it is computationally more demanding. In this paper, we compare the EKF and PF performance in the dual Bayesian estimation of battery state and parameters, with particular reference to lithium batteries, showing that PF is attractive, especially in the presence of inaccurate battery models
Deep learning processing of remotely sensed multi-spectral images
No full textThis chapter covers the most recent advancements in deep learning approaches tailored to multi-spectral remotely sensed images. Multi-spectral imaging conveys detailed information across several wavelengths, allowing for better environmental monitoring, precision agriculture, urban planning, and disaster management. The ability of deep learning-based approaches to extract complex patterns and features holds prospective in this domain. We specifically explore the challenges that these images give, including disparities in spatial resolution, spectral variability, and a lack of labelled data, while concurrently looking at cutting-edge deep learning-based algorithms and learning techniques specifically designed to deal with them. By summarizing current developments and outlining future research objectives, this chapter serves as a valuable resource for academics and professionals seeking to leverage deep learning for multi-spectral remote sensing image analysis
- …
