Welcome to the BUT Digital Library - an institutional repository operated by the Central Library on the DSpace system.
Do you want to deposit your article or preceedings into Digital Library? It is very simple. You can find all the information in the manual published online on BUT Portal of libraries.
Central Library supports open access to scientific publishing - Open Access.
You can also request for grant for open publishing from Open Access Fund You can find more information OA fund web page.
Into the Digital Library is integrated citation manager Citace PRO. It will allow you to easily create a bibliographic citation or save a record in the manager.
Communities in DSpace
Select a community to browse its collections.
Recent Submissions
Item
A Streaming Dataflow Accelerator for Sparse SVM Kernel Computation in Hyperspectral Image Classification
(Radioengineering Society, 2025-09) Shabarinath, B. B.; Muralidhar, P.
Hyperspectral images (HSI) provide extensive spectral information but their high dimensionality and redundancy create substantial challenges for computation and storage while increasing energy demands. The proposed solution combines sparse dictionary learning with Field Programmable gate Array (FPGA)-accelerated Sparse matrix vector multiplication (SpMV) operations and Support Vector Machine (SVM) training to tackle these issues. Spatial patches and spectral blocks partition HSI to enable the extraction of compact discriminative sparse features through the use of a learned sub-dictionary. In contrast to deep learning frameworks which demand large training datasets and generate significant computational overhead, the SVM-based approach achieves efficient real-time training and adaptation. The FPGA accelerator executes intensive SpMV operations through dynamic load balancing. We tested our approach with four varied HSI datasets gathered from aerial and UAV systems as well as terrestrial platforms on the PYNQ-Z2 board. Our design reaches classification accuracies between 98.65% and 99.95% across datasets including Indian Pines,AVRIS-NG, Cubert-UAV, Cubert-Terrestrial with per-pixel classification times below 7 us and inference times up to 36x faster than optimized software baselines which remain under typical sensor acquisition times. The strategy requires less than 0.24 W of on-chip power at maximum load which makes it ideal for deployment on satellites or UAVs. The proposed method outperforms existing FPGA-based SVM architectures in classification accuracy and throughput while enabling on-device incremental learning which makes it ideal for analyzing hyperspectral images in real-time.
Item
Self-Supervised Learning Driven Cross-Domain Feature Fusion Network for Hyperspectral Image Classification
(Radioengineering Society, 2025-09) Fang, Q.; Zhao, Y.; Wang, J.; Zhang, L.
Hyperspectral image (HSI) classification faces significant challenges due to the high cost of acquiring labeled samples. To mitigate this, we propose SSCF-Net, a novel self-supervised learning driven cross-domain feature fusion Network. SSCF-Net uniquely leverages readily available labeled natural images (source domain) to aid HSI (target domain) classification by transfer learning. Specifically, we employ rotation-based self-supervision in the source domain to learn transferable features, which are then transferred to the HSI domain. Within SSCF-Net, we effectively fuse local and global features: local features are extracted by a jointly trained module combining VGG and two-dimensional long short-term memory networks (TD-LSTM) networks, while global features capturing long-range dependencies are learned via a Transformer model. Crucially, in the HSI domain, we further employ contrastive learning as a self-supervised strategy to maximally utilize the limited labeled data. Extensive experiments on three benchmark HSI datasets (Salinas, Indian Pines, WHU-Hi-LongKou) demonstrate that SSCF-Net significantly outperforms existing methods, validating its effectiveness in addressing the label scarcity problem. The code is available at https://github.com/6pangbo/SSCF-Net.
Item
LVRT Strategy Considering Reactive Power Support and Fluctuating Power Suppression for Photovoltaic Application
(Radioengineering Society, 2025-09) Zhang, Z.; Liu, Q.; Wang, Y.
Addressing the insufficient negative sequence dynamic reactive current support and power doubling oscillations present in conventional low-voltage ride through control mechanisms for photovoltaic inverters, this study designs a q-axis command for both positive and negative sequence currents in accordance with recent regulatory requirements for photovoltaic grid integration technologies. The d-axis commands for positive and negative sequence currents are computed to effectively attenuate second harmonic fluctuations in active power output. The proposed approach establishes equilibrium between inverter current carrying capacity and oscillation mitigation, thereby concurrently enhancing dynamic reactive current support for both sequence components while diminishing power doubling fluctuations. The short-circuit current characteristics of photovoltaic installations are examined utilizing this enhanced low-voltage ride through control methodology. Comparative analysis between the suggested approach and current low-voltage ride through control techniques was conducted via simulation models, confirming both the efficiency of the proposed method and the accuracy of analytical expressions for short-circuit current.
Item
An Enhanced Noise Removal-based SAR Image Recognition Using DnCNN and Wavelet Transform
(Radioengineering Society, 2025-09) Choi, Y.; Kim, G.; Kim, B.; Kim, S.
This paper presents an enhanced method for noise removal and target detection in Synthetic Aperture Radar (SAR) images using a Denoising Convolutional Neural Network (DnCNN) combined with wavelet trans¬form. Unlike conventional method, the proposed frame¬work focuses on remove the Speckle Noise through residu¬al learning and wavelet transform. The DnCNN architecture, consisting of 29 layers, efficiently removes noise while preserving high-frequency image features. The integration of wavelet transform further enhances noise removal and feature preservation. Experimental results demonstrate that the recognition rate of the proposed method improves by about 94% compared to original method under 10 dB Speckle Noise conditions. This method outperforms conventional algorithm in SAR image pro¬cessing, making it highly suitable for applications in noisy environments.
Item
A Novel Design of a Low-loss and Low-cost Ku-Band Bandpass Filter for VSAT Applications
(Radioengineering Society, 2025-09) Ta, M. V.; Phuong, K. K.; Trong, N. D.; Linh, N. T.; Huong, T. T. T.; Hung, N. T.; Manh, L. D.
This paper proposes a novel method to design low-loss and low-cost Ku-band bandpass filters for VSAT applications based on substrate-integrated-waveguide technology. Narrow bandpass filters employed high-order resonant mode TE301 exhibited high selectivity. However, its bandwidth is not enough for VSAT applications. In this paper, we proposed a method to widen the bandwidth of narrow-band filters to meet the bandwidth requirement of VSAT applications. This approach maintains high selectivity while still achieves low insertion loss. The proposed filter was fabricated using a low-cost material. Measurement shows a good agreement with simulated results. Mid-band measured insertion loss and return loss were -1.8 dB and -19.4 dB, respectively. Such low losses were obtained owing to taking advantage of a high-quality factor of high-order mode TE301 of oversized rectangular cavities