ZHAO LIYUAN

Dr. ZHAO LIYUAN

National University of Defense Technology, Changsha, Guangdong, China

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
10 August 2021 in Applied Sciences
Verified
Journal Article
Open access
PDF file
JLJian Long
Jian Long
YPYuanxi Peng
Yuanxi Peng
TZTong Zhou
Tong Zhou
ZLZHAO LIYUAN
ZHAO LIYUAN
JLJun Li
Jun Li
Fusion low-resolution hyperspectral images (LR-HSI) and high-resolution multispectral images (HR-MSI) are important methods for obtaining high-resolution hyperspectral images (HR-HSI). Some hyperspectral image fusion application areas have strong real-time requirements for image fusion, and a fast fusion method is urgently needed. This paper proposes a fast and stable fusion method (FSF) based on matrix factorization, which can largely reduce the computational workloads of image fusion to achieve fast and efficient image fusion. FSF introduces the Moore–Penrose inverse in the fusion model to simplify the estimation of the coefficient matrix and uses singular value decomposition (SVD) to simplify the estimation of the spectral basis, thus significantly reducing the computational effort of model solving. Meanwhile, FSF introduces two multiplicative iterative processes to optimize the spectral basis and coefficient matrix to achieve stable and high-quality fusion. We have tested the fusion method on remote sensing and ground-based datasets. The experiments show that our proposed method can achieve the performance of several state-of-the-art algorithms while reducing execution time to less than 1% of such algorithms.

25 October 2022 in Photonics
Verified
Journal Article
Open access
PDF file
ZLZHAO LIYUAN
ZHAO LIYUAN
JZJianghua Zhang
Jianghua Zhang
LHLei Huang
Lei Huang
YPYuanxi Peng
Yuanxi Peng
KYKe Yin
Ke Yin
HNHongxiao Niu
Hongxiao Niu
Due to the nonlinear and aliasing effects, the sub-Nyquist photonic receiver for radio frequency (RF) signals with large instantaneous bandwidth suffers limited dynamic range and noise performance. We designated a deep residual network (Resnet) to realize adaptive linearization across 40 GHz bandwidth. In contrast to conventional linearization methods, the deep learning method achieves the suppression of multifactorial spurious distortions and the noise floor simultaneously. It does not require an accurate calculation of the nonlinear transfer function or prior signal information. The experiments demonstrated that the proposed Resnet could improve the spur-free dynamic range (SFDR) and the signal-to-noise ratio (SNR) significantly by testing with single-tone signals, dual-tone signals, wireless communication signals, and modulated radar signals.