A5076077485- Advanced Steganography and Watermarking Techniques
- Digital Media Forensic Detection
- Adversarial Robustness in Machine Learning
- Advanced SAR Imaging Techniques
- Bacillus and Francisella bacterial research
- Integrated Circuits and Semiconductor Failure Analysis
- Geophysical Methods and Applications
- Video Coding and Compression Technologies
- Medical Image Segmentation Techniques
- Chaos-based Image/Signal Encryption
- Underwater Acoustics Research
- Image Processing Techniques and Applications
- Advanced Image Processing Techniques
- Infrared Target Detection Methodologies
- Recycling and utilization of industrial and municipal waste in materials production
- Structural Health Monitoring Techniques
- Medical Imaging and Analysis
- Anomaly Detection Techniques and Applications
- Fault Detection and Control Systems
- Image Processing and 3D Reconstruction
- Spectroscopy Techniques in Biomedical and Chemical Research
- Direction-of-Arrival Estimation Techniques
- Speech and Audio Processing
- Ultrasonics and Acoustic Wave Propagation
- Advanced Neural Network Applications
National University of Defense Technology
2022-2024
Quality and Reliability (Greece)
2022
South China University of Technology
2020-2021
Current video steganography operates with either the decoded frame images or compression coding parameters, which could cause quality degradation of reconstructed frames. In this letter, by exploiting advanced motion vector prediction (AMVP) technique High Efficiency Video Coding (HEVC) standard, we propose a non-degraded adaptive steganographic approach for H.265/HEVC videos. The index value in candidate list unit (PU) is used embedding. Experimental results demonstrate superiority proposed...
Recently, deep unfolding networks have been widely used in direction of arrival (DOA) estimation because their improved accuracy and reduced computational cost. However, few considered the existence a nested array (NA) with off-grid DOA estimation. In this study, we present sparse Bayesian learning (DSBL) network to solve problem. We first establish signal model for NA. Then, transform output into real domain neural networks. Finally, construct train DSBL determine on-grid spatial spectrum...
Generally, the purpose of a steganalysis algorithm is to establish presence secret messages in stego data. However, quantitative steganalyzers can reveal more information about communication by estimating exact volume embedded messages. Quantitative crucial step for breaking codes many practical scenarios. This work concerns videos. Most video steganographical algorithms embed modifying values motion vector compressed domain. We propose general framework constructing that are able detect...
The paper focuses on the multi-azimuth interpolation task of inverse synthetic aperture radar (ISAR) images for aircraft targets and complements incomplete ISAR image datasets. automatic target recognition (ATR) has been widely applied in remote sensing many fields. However, imaging process is more challenging when compared to capturing optical SAR data, which reduces accuracy generalization performance ATR system. Therefore, this paper, we leverage existing limited data achieve autonomous...
Adversarial examples (AEs) bring increasing concern on the security of deep-learning-based synthetic aperture radar (SAR) target recognition systems. SAR AEs with perturbation constrained to vicinity have been recently in spotlight due physical realization prospects. However, current adversarial detection methods generally suffer severe performance degradation against region-constrained perturbation. To solve this problem, we treated as low-probability samples incompatible clean dataset....
It is well established that using the selection channel, probabilities with which elements in cover are modified during message embedding, would improve performance of steganalysis. Most video steganographical algorithms embed secret messages compressed domain by modifying motion vectors having less impact on visual quality, can be considered as a form channel. Recently, deep neural networks have been rapidly developed for multimedia Although there some selection-channel-aware image...
Inverse synthetic aperture radar (ISAR) image offers geometric and structural characteristics information of target objects. Thus, It is an important research topic to recognize targets based on ISAR images. imaging the advantages all-day, all-weather, ultra-long- distance imaging; however, quality affected by attitude angle, defocusing noise, resolution other factors, resulting in inferior recognition performance. In contrast, optical images require more stringent conditions, but they...
In recent years, radar automatic target recognition tech-nology based on high-resolution range profile has received widespread attention. Most deep learning-based methods do not make use of the temporal dependence. Given this issue, we introduce Transformer model for HRRP tasks in work. However, existing transformer models need to manually design position embedding mode, and lack inductive hypothesis. Therefore, propose a embedding-free combined with CNN. Specifically, self attention module...
Nowadays, deep learning techniques are extensively applied in the field of automatic target recognition (ATR) for radar images. However, existing data-driven approaches frequently ignore prior knowledge target, leading to a lack interpretability and poor performance trained models. To address this issue, we first integrate structural attributes into training process an ATR model, providing both category information at dataset level. Specifically, propose Structural Attribute Injection (SAI)...
In this paper, we propose a convolutional neural network (CNN) with multiscale window self-attention mechanism for radar high-resolution range profile (HRRP) target recognition task. Specifically, take one-dimensional CNN (1-D CNN) to extract shallow feature and fully excavate the rich local structural features of HRRP data. Then, utilize capture regional difference The proposed module divides obtained by into equal width bands through sliding windows. Multi-level different regions can be...
Adversarial attack and adversarial detection have become a hot issue in the field of deep learning based image forensics. However, current researches mainly focus on optical images. Synthetic aperture radar (SAR) images are quite different from both imaging mechanism data structure. This paper aims to study for SAR Firstly, we analyze distribution characteristics examples (AEs) output space feature by transferring attacks. In order match digital perturbation with scattering energy target,...
Adversarial examples (AEs) have become a critical security concern for intelligent synthetic aperture radar (SAR) target recognition system. Current defense methods design certain distance metric to characterize the difference between AEs and natural samples in feature space, but expose severe performance degradation against SAR with small perturbation scale. By exploiting Siamesed sample augmentation wised contrastive regularization, we propose joint cross-entropy training method endow test...
Adversarial attack and adversarial detection have become a hot issue in the field of deep learning based image forensics. However, current researches mainly focus on optical images. Synthetic aperture radar (SAR) images are quite different from both imaging mechanism data structure. This paper aims to study for SAR Firstly, we analyze distribution characteristics examples (AEs) output space feature by simply transferring attacks. In order match digital perturbation with scattering energy...