Yazar "Kayacan, Yavuz Emre" seçeneğine göre listele

Listeleniyor 1 - 4 / 4
  • Küçük Resim Yok
    Öğe
    A Vision-Transformer-Based Approach to Clutter Removal in GPR: DC-ViT
    (Ieee-Inst Electrical Electronics Engineers Inc, 2024) Kayacan, Yavuz Emre; Erer, Isin
    Since clutter encountered in ground-penetrating radar (GPR) systems deteriorates the performance of target detection algorithms, clutter removal is an active research area in the GPR community. In this letter, instead of convolutional neural network (CNN) architectures used in the recently proposed deep-learning-based clutter removal methods, we introduce declutter vision transformers (DC-ViTs) to remove the clutter. Transformer encoders in DC-ViT provide an alternative to CNNs which has limitations to capture long-range dependencies due to its local operations. In addition, the implementation of a convolutional layer instead of multilayer perceptron (MLP) in the transformer encoder increases the capturing ability of local dependencies. While deep features are extracted with blocks consisting of transformer encoders arranged sequentially, losses during information flow are reduced using dense connections between these blocks. Our proposed DC-ViT was compared with low-rank and sparse methods such as robust principle component analysis (RPCA), robust nonnegative matrix factorization (RNMF), and CNN-based deep networks such as convolutional autoencoder (CAE) and CR-NET. In comparisons made with the hybrid dataset, DC-ViT is 2.5% better in peak signal-to-noise ratio (PSNR) results than its closest competitor. As a result of the tests, we conducted using our experimental GPR data, and the proposed model provided an improvement of up to 20%, compared with its closest competitor in terms of signal-to-clutter ratio (SCR).
  • Küçük Resim Yok
    Öğe
    Autoencoder Guided Low-Rank Approximation Approach for Clutter Removal in GPR Images
    (Institute of Electrical and Electronics Engineers Inc., 2024) Kayacan, Yavuz Emre; Erer, Isin
    The performance of low-rank and sparse decomposition (LRSD) based clutter removal methods which are widely used in GPR systems depends heavily on the regularization parameter. This study proposes a A parameter-free low-rank approach. The low-rank component recovered by an autoencoder (AE) network is subtracted from the raw image to provide a clutter-free image. Simulation and experimental results validate the superiority of the proposed method compared to the low-rank approach Nonnegative Matrix Factorization (NMF) as well as other LRSD methods: Robust Principal Component Analysis (RPCA), Robust NMF (RNMF), and Robust Autoencoder (RAE). © 2024 IEEE.
  • Küçük Resim Yok
    Öğe
    Combining Image Restoration with Deep Detectors for Improved Target Detection in GPR
    (Institute of Electrical and Electronics Engineers Inc., 2024) Kayacan, Yavuz Emre; Özdoǧan, Kaan; Çolak, Mehmet Utku; Aydin, Özae; Turan, Ahmet Enes; Erer, Isin
    The target detection performance of GPR systems suffers from the noisy and partially lost data encountered in field studies. Since detection may benefit from the restoration of the data prior to detection procedure, a two-step approach is proposed. Firstly, the data is either denoised or recovered with networks appropriate to the nature of the task, then detection is performed by YOLOv5 or YOLOv10. Detection results obtained with the proposed approach confirm performance increase in terms of MAP values compared to the sole use of YOLOv5 or YOLOv10. © 2024 IEEE.
  • Küçük Resim Yok
    Öğe
    Enhancing Deep Learning Networks Performance By Using B-Spline Activation Functions for Clutter Removal in GPR
    (Institute of Electrical and Electronics Engineers Inc., 2024) Kayacan, Yavuz Emre; Erer, Isin
    Traditional clutter removal methods struggle with complex clutter and multiple targets in Ground-Penetrating Radar images. This study proposes using B-spline activation functions in the deep-learning models to improve clutter removal in GPR. Experimental results demonstrate that B-spline-enhanced models outperform their ReLU-based counterparts, with improvements of up to 2.20% in PSNR, 0.035% in MS-SSIM, and 15.83% in SCR, showcasing their potential for real-world GPR applications. © 2024 IEEE.