| | SLO | ENG | Piškotki in zasebnost

Večja pisava | Manjša pisava

Iskanje po katalogu digitalne knjižnice Pomoč

Iskalni niz: išči po
išči po
išči po
išči po
* po starem in bolonjskem študiju

Opcije:
  Ponastavi


1 - 10 / 25
Na začetekNa prejšnjo stran123Na naslednjo stranNa konec
1.
2.
Cross-Hole GPR for Soil Moisture Estimation Using Deep Learning
Blaž Pongrac, Dušan Gleich, Marko Malajner, Andrej Sarjaš, 2023, izvirni znanstveni članek

Opis: This paper presents the design of a high-voltage pulse-based radar and a supervised data processing method for soil moisture estimation. The goal of this research was to design a pulse-based radar to detect changes in soil moisture using a cross-hole approach. The pulse-based radar with three transmitting antennas was placed into a 12 m deep hole, and a receiver with three receive antennas was placed into a different hole separated by 100 m from the transmitter. The pulse generator was based on a Marx generator with an LC filter, and for the receiver, the high-frequency data acquisition card was used, which can acquire signals using 3 Gigabytes per second. Used borehole antennas were designed to operate in the wide frequency band to ensure signal propagation through the soil. A deep regression convolutional network is proposed in this paper to estimate volumetric soil moisture using time-sampled signals. A regression convolutional network is extended to three dimensions to model changes in wave propagation between the transmitted and received signals. The training dataset was acquired during the period of 73 days of acquisition between two boreholes separated by 100 m. The soil moisture measurements were acquired at three points 25 m apart to provide ground truth data. Additionally, water was poured into several specially prepared boreholes between transmitter and receiver antennas to acquire additional dataset for training, validation, and testing of convolutional neural networks. Experimental results showed that the proposed system is able to detect changes in the volumetric soil moisture using Tx and Rx antennas.
Ključne besede: ground penetrating radar, cross-hole, L-band, deep learning, convolutional neural network, soil moisture estimation
Objavljeno v DKUM: 03.04.2024; Ogledov: 76; Prenosov: 4
.pdf Celotno besedilo (3,22 MB)
Gradivo ima več datotek! Več...

3.
Prediction of the form of a hardened metal workpiece during the straightening process
Tadej Peršak, Jernej Hernavs, Tomaž Vuherer, Aleš Belšak, Simon Klančnik, 2023, izvirni znanstveni članek

Opis: In industry, metal workpieces are often heat-treated to improve their mechanical properties, which leads to unwanted deformations and changes in their geometry. Due to their high hardness (60 HRC or more), conventional bending and rolling straightening approaches are not effective, as a failure of the material occurs. The aim of the research was to develop a predictive model that predicts the change in the form of a hardened workpiece as a function of the arbitrary set of strikes that deform the surface plastically. A large-scale laboratory experiment was carried out in which a database of 3063 samples was prepared, based on the controlled application of plastic deformations on the surface of the workpiece and high-resolution capture of the workpiece geometry. The different types of input data, describing, on the one hand, the performed plastic surface deformations on the workpieces, and on the other hand the point cloud of the workpiece geometry, were combined appropriately into a form that is a suitable input for a U-Net convolutional neural network. The U-Net model’s performance was investigated using three statistical indicators. These indicators were: relative absolute error (RAE), root mean squared error (RMSE), and relative squared error (RSE). The results showed that the model had excellent prediction performance, with the mean values of RMSE less than 0.013, RAE less than 0.05, and RSE less than 0.004 on test data. Based on the results, we concluded that the proposed model could be a useful tool for designing an optimal straightening strategy for high-hardness metal workpieces. Our results will open the doors to implementing digital sustainability techniques, since more efficient handling will result in fewer subsequent heat treatments and shorter handling times. An important goal of digital sustainability is to reduce electricity consumption in production, which this approach will certainly do.
Ključne besede: sustraightening process, hardened workpiece, manufacturing, U-Net convolutional neural network, modeling, point cloud, digital sustainability
Objavljeno v DKUM: 02.04.2024; Ogledov: 73; Prenosov: 11
.pdf Celotno besedilo (10,52 MB)
Gradivo ima več datotek! Več...

4.
Influence of Al2O3 nanoparticles addition in ZA-27 alloy-based nanocomposites and soft computing prediction
Aleksandar Vencl, Petr Svoboda, Simon Klančnik, Adrian But, Miloš Vorkapić, Marta Harničárová, Blaža Stojanović, 2023, izvirni znanstveni članek

Opis: Three different and very small amounts of alumina (0.2, 0.3 and 0.5 wt. %) in two sizes (approx. 25 and 100 nm) were used to enhance the wear characteristics of ZA-27 alloy-based nanocomposites. Production was realised through mechanical alloying in pre-processing and compocasting processes. Wear tests were under lubricated sliding conditions on a block-on-disc tribometer, at two sliding speeds (0.25 and 1 m/s), two normal loads (40 and 100 N) and a sliding distance of 1000 m. Experimental results were analysed by applying the response surface methodology (RSM) and a suitable mathematical model for the wear rate of tested nanocomposites was developed. Appropriate wear maps were constructed and the wear mechanism is discussed in this paper. The accuracy of the prediction was evaluated with the use of an artificial neural network (ANN). The architecture of the used ANN was 4-5-1 and the obtained overall regression coefficient was 0.98729. The comparison of the predicting methods showed that ANN is more efficient in predicting wear.
Ključne besede: ZA-27 alloy, Al2O3 nanoparticles, nanocomposites, wear, response surface methodology, artificial neural network
Objavljeno v DKUM: 20.03.2024; Ogledov: 105; Prenosov: 4
.pdf Celotno besedilo (14,10 MB)
Gradivo ima več datotek! Več...

5.
Despeckling of SAR Images Using Residual Twin CNN and Multi-Resolution Attention Mechanism
Blaž Pongrac, Dušan Gleich, 2023, izvirni znanstveni članek

Opis: The despeckling of synthetic aperture radar images using two different convolutional neural network architectures is presented in this paper. The first method presents a novel Siamese convolutional neural network with a dilated convolutional network in each branch. Recently, attention mechanisms have been introduced to convolutional networks to better model and recognize features. Therefore, we propose a novel design for a convolutional neural network using an attention mechanism for an encoder–decoder-type network. The framework consists of a multiscale spatial attention network to improve the modeling of semantic information at different spatial levels and an additional attention mechanism to optimize feature propagation. Both proposed methods are different in design but they provide comparable despeckling results in subjective and objective measurements in terms of correlated speckle noise. The experimental results are evaluated on both synthetically generated speckled images and real SAR images. The methods proposed in this paper are able to despeckle SAR images and preserve SAR features.
Ključne besede: synthetic aperture radar, speckle, speckle suppression, despeckling, deep learning, convolutional neural network
Objavljeno v DKUM: 21.02.2024; Ogledov: 125; Prenosov: 10
.pdf Celotno besedilo (13,70 MB)
Gradivo ima več datotek! Več...

6.
7.
Frequency range optimization for continuous wave Terahertz imaging
Blaž Pongrac, Andrej Sarjaš, Dušan Gleich, izvirni znanstveni članek

Opis: S krajšimi valovnimi dolžinami kot mikrovalovi in večjo globino prodora v material kot infrardeča svetloba, valovi v TeraHertz-nem (THz) spektru ponujajo edinstvene možnosti testiranja materialov. THz tehnologija ponuja neinvazivna in nedestruktivna testiranja v obliki spektroskopije in slikanja. Najbolj uporabljeni sistemi za THz slikanje so sistemi spektroskopije v časovni domeni. Vendar sistemi spektroskopije frekvenčne domene ponujajo odlično frekvenčno ločljivost in so primerni za biomedicinske aplikacije. THz-no slikanje na podlagi spektroskopije v frekvenčnem prostoru je časovno kompleksno in ima pomanjkljivosti zaradi napak pri generiranju THz valov. V tem članku je predstavljen nov princip enodimenzionalnega zajemanja s THz valovi. Predlagana optimizacija frekvenčnega območja temelji na konvolucijski nevronski omreži. Predstavljena je frekvenčna optimizacija za določitev optimalnega frekvenčnega območja za zajem podatkov. Optimalno frekvenčno območje ali pasovna širina morata biti dovolj široka za učinkovito zaznavanje faze in morata biti na presečišču več spektralnih odtisov v opazovanem mediju. Presek spektralnih odtisov je ocenjen z uporabo predlaganega algoritma za optimizacijo frekvenčnega območja, ki temelji na konvolucijski nevronski mreži in algoritmu za občutljivost okluzije. Predlagani algoritem izbira samodejno najobčutljivejši frekvenčni pas THz spektra in omogoča zelo hitre zajeme za pregled in klasifikacijo objektov.
Ključne besede: terahertz, spectroscopy, imaging, convolutional neural network, occlusion sensitivity, optimization
Objavljeno v DKUM: 07.12.2023; Ogledov: 267; Prenosov: 9
.pdf Celotno besedilo (8,82 MB)
Gradivo ima več datotek! Več...

8.
Naive prediction of protein backbone phi and psi dihedral angles using deep learning
Matic Broz, Marko Jukič, Urban Bren, 2023, izvirni znanstveni članek

Opis: Protein structure prediction represents a significant challenge in the field of bioinformatics, with the prediction of protein structures using backbone dihedral angles recently achieving significant progress due to the rise of deep neural network research. However, there is a trend in protein structure prediction research to employ increasingly complex neural networks and contributions from multiple models. This study, on the other hand, explores how a single model transparently behaves using sequence data only and what can be expected from the predicted angles. To this end, the current paper presents data acquisition, deep learning model definition, and training toward the final protein backbone angle prediction. The method applies a simple fully connected neural network (FCNN) model that takes only the primary structure of the protein with a sliding window of size 21 as input to predict protein backbone φ and ψ dihedral angles. Despite its simplicity, the model shows surprising accuracy for the φ angle prediction and somewhat lower accuracy for the ψ angle prediction. Moreover, this study demonstrates that protein secondary structure prediction is also possible with simple neural networks that take in only the protein amino-acid residue sequence, but more complex models are required for higher accuracies.
Ključne besede: protein structure prediction, backbone dihedral angles, deep neural network, fully connected neural network, FCNN, protein secondary structure prediction
Objavljeno v DKUM: 01.12.2023; Ogledov: 191; Prenosov: 21
.pdf Celotno besedilo (3,60 MB)
Gradivo ima več datotek! Več...

9.
Development of a methodology to calibrate a pedestrian microsimulation model : doctoral dissertation
Chiara Gruden, 2022, doktorska disertacija

Opis: Walking, as a mode of transport, is becoming widespread, in a world, where urban conglomerates are broadening and becoming denser. Modern lifestyle trends on a side, and eco-friendly policies on the other, push people into walking habits, increasing the need for a suitable, attractive, accessible, connected and safe walking infrastructure. To reach such a result, it is necessary to understand, what are the needs of the users of this infrastructure, taking into consideration the behavioral specificities and the safety needs of pedestrians. In this process pedestrian microsimulation models, surrogate safety techniques, and technologies able to measure specific traits of pedestrian dynamics play a central role. The firsts allow to reproduce repeatedly in a virtual environment a specific infrastructure and to study the response of pedestrians. Nevertheless, to be accurate and efficient, they need to go through long and tedious calibration and validation processes, that are often seen as an important limitation by technicians. Surrogate safety techniques are methods, that are based on the concept, that it is possible to predict the safety level of a location, using near accidents. The main advantage of such techniques is that they are proactive. Till this moment, these techniques have been mainly applied to on-field measurements and are primarily centered on motorized road users. Less interest has been shown for vulnerable road users, especially for pedestrians, who have been less extensively studied. Finally, an element that could highly affect pedestrian safety is their reaction time. Nevertheless, its measurement has long been a big issue. Eye-tracking technology could be one of the solutions, allowing to analyze the directions and objects fixated by pedestrians. These listed issues are also the topics that are addressed by this research work. Focusing on the study of the action of pedestrians while crossing the road on an unsignalized crosswalk set on a roundabout entry leg, the dissertation thesis aims at studying the crossing time, reaction time and surrogate safety aspects typical of pedestrians at the recalled location. The main purpose of the research work is to develop a methodology to calibrate pedestrian Social Force Model at a selected location, using a specifically formulated neural network as a tool to fine-tune model's behavioral parameters. Eight parameters have been chosen to be fine-tuned, five of those are related to pedestrian behavior and three of them are related to car-following behavior. After the selection of input parameters, a feedforward network has been formulated. Its application in the framework of the whole calibration process has brought to considerably positive results, finding a combination of input parameters that improved the performance of the microsimulation model of 37 % in comparison to the default one. The outputs of the calibrated model have been used to calculate three measures of surrogate safety, and also in this case results demonstrated an improvement in the calculation of surrogate safety measures when using the calibrated outcomes in comparison to their calculation on the “default” model outputs. Finally, reaction time measurement and prediction have been addressed by the thesis, in order to be able to describe pedestrian crossing action in its completeness. Quantitative eye-tracking outputs have been the starting point for the calculation of pedestrian reaction time at different locations, and they allowed to create a database of behavioral, geometric, regulatory and flow characteristics, which was the foundation for the formulation of a new prediction model for pedestrian reaction time. The prediction model, which consists of a cascade-correlation neural network, gave a good response to the learning and generalization steps, turning a 74 % correlation between the measured reaction time values and the predicted ones, and being able to follow the variability of these values.
Ključne besede: pedestrian, microsimulation model, calibration, neural network, surrogate safety indicators, reaction time.
Objavljeno v DKUM: 03.10.2022; Ogledov: 717; Prenosov: 93
.pdf Celotno besedilo (5,93 MB)

10.
ARM-Based Video Intercom System with Next-Gen Human Presence Detection using Deep Learning : magistrsko delo
Mario Gavran, 2022, magistrsko delo

Opis: This master's thesis presents an advanced video system with human presence detection based on deep learning and an ARM microcontroller. The objective of the thesis is to develop a system that works as a smart video intercom, which could be installed, e.g. on the entrance door, and autonomously alert the owner that a guest is in front of the door. The main goal is to use an AI algorithm, namely the neural network model on a constrained device, such as an ARM microcontroller, as their main advantage is lower power consumption and cost. The thesis also describes commonly used methods to reduce the power and memory footprint and to implement and accelerate the deep learning algorithms more effectively. Further, the most notable deep learning hardware and some general platforms are described in more detail. The thesis also presents the development of a human presence detection system based on an ARM microcontroller, VGA camera, and LCD, where Tensorflow Lite Micro, an open-source C++ framework for deploying deep learning models to embedded platforms and a pre-trained neural network model for person presence detection are used.
Ključne besede: TensorFlow Lite Micro, Video intercom system, ARM Cortex-M microcontroller, Human presence detection, Neural network
Objavljeno v DKUM: 08.07.2022; Ogledov: 658; Prenosov: 52
.pdf Celotno besedilo (7,86 MB)

Iskanje izvedeno v 0.85 sek.
Na vrh
Logotipi partnerjev Univerza v Mariboru Univerza v Ljubljani Univerza na Primorskem Univerza v Novi Gorici