A5015425431- Image Enhancement Techniques
- Infrared Target Detection Methodologies
- Advanced Image Fusion Techniques
- Video Surveillance and Tracking Methods
- Advanced Vision and Imaging
- Remote-Sensing Image Classification
- Advanced Image Processing Techniques
- Advanced Measurement and Detection Methods
- Optical measurement and interference techniques
- Advanced SAR Imaging Techniques
- Wireless Signal Modulation Classification
- Geophysical Methods and Applications
- Human Pose and Action Recognition
- Advanced Chemical Sensor Technologies
- Thermography and Photoacoustic Techniques
- Advanced Neural Network Applications
Hefei University of Technology
2020-2024
Single-frame infrared small target detection is still a challenging task due to the complex background and unobvious structural characteristics of targets. Recently, convolutional neural networks (CNN) began appear in field have been widely used for excellent performance. However, existing CNN-based methods mainly focus on local spatial features while ignoring long-range contextual dependencies between targets backgrounds. To capture global context-aware information, we propose fusion...
The fusion of infrared intensity and polarization images can generate a single image with better visible perception more vital information. Existing methods based on convolutional neural network (CNN), local feature extraction, have the limitation fully exploiting salient target features polarization. In this Letter, we propose transformer-based deep to improve performance fusion. Compared existing CNN-based methods, our model encode long-range obtain global contextual information using...
Typical infrared polarization image fusion aims to integrate background details in the intensity and salient target degree of linear (DoLP). Many methods show advanced network architecture, but few works can form effective feature representations for differences prior distributions DoLP, interference DoLP with noise makes more challenging. This paper employs a learned low-rank decomposition model extract containing sparse features targets DoLP. To reduce interference, we design module based...
Light field cameras record not only the spatial information of observed scenes but also directions all incoming light rays. The and angular implicitly contain geometrical characteristics such as multi-view or epipolar geometry, which can be exploited to improve performance depth estimation. An Epipolar Plane Image (EPI), unique 2D spatial-angular slice field, contains patterns oriented lines. slope these lines is associated with disparity. Benefiting from this property EPIs, some...
Infrared polarization image fusion integrates intensity and information, producing a fused that enhances visibility captures crucial details. However, in complex environments, imaging is susceptible to noise interference. Existing methods typically use the infrared (S0) degree of linear (DoLP) images for but fail consider interference, leading reduced performance. To cope with this problem, we propose method based on salient prior, which extends DoLP by angle (AoP) introduces distance (PD)...
Recent advancements in road detection using infrared polarization imaging have shown promising results. However, existing methods focus on refined network structures without effectively exploiting mechanisms for enhanced detection. The scarcity of datasets also limits the performance these methods. In this Letter, we present a denoising diffusion model aimed at improving images. This achieves effective integration intensity and information through forward reverse processes. Furthermore,...
Exploiting light field data makes it possible to obtain dense and accurate depth map. However, synthetic scenes with limited disparity range cannot contain the diversity of real scenes. By training in data, current learning based methods do not perform well In this paper, we propose a self-supervised framework for estimation. Different from existing end-to-end using label per pixel, our approach implements network by estimating EPI shift after refocusing, which extends epipolar lines. To...
The RGB-D cross-modal person re-identification task is a branch of the task, aiming to match pedestrian images between RGB and depth modalities. Its main challenge lies in significant modality difference images. Due limited research attention on this there lack large-scale, high-quality training data, making existing methods struggle capture effective correlational information This paper proposes novel feature embedding expansion network for re-identification. introduces perturbations...
Radar radiation source sorting and identification methods based on in-depth learning have received considerable attention in recent years, but there are time resource constraints practice. To solve this problem, paper presents a method of radar signal meta-learning. Data samples for generated by acquiring the three-dimensional entropy, image coefficient, high-order cumulants, bandwidth carrier frequency characteristics signal. A classification model adaptive channel is constructed....