A5101567713- Advanced Image Processing Techniques
- Video Surveillance and Tracking Methods
- Advanced Vision and Imaging
- Robotics and Sensor-Based Localization
- Image Enhancement Techniques
- Advanced Image Fusion Techniques
- Advanced Optical Sensing Technologies
- Image Processing Techniques and Applications
- Image and Signal Denoising Methods
- Optical measurement and interference techniques
- Image and Object Detection Techniques
- Advanced Image and Video Retrieval Techniques
- 3D Surveying and Cultural Heritage
- Robotic Path Planning Algorithms
- Remote Sensing and LiDAR Applications
- Advanced Malware Detection Techniques
- Remote-Sensing Image Classification
- Underwater Acoustics Research
- Anomaly Detection Techniques and Applications
- Visual Attention and Saliency Detection
- Hydraulic and Pneumatic Systems
- Robotic Locomotion and Control
- Impact of Light on Environment and Health
- Structural Load-Bearing Analysis
- Topic Modeling
UNSW Sydney
2025
ACT Government
2025
Beijing Electronic Science and Technology Institute
2024
Inner Mongolia University of Technology
2023
Xi'an Jiaotong University
2015-2021
Tianjin University
2018-2021
Harbin Engineering University
2020
Southern University of Science and Technology
2020
Seoul National University
2019
Tongji University
2017-2018
In this paper, we propose a Multi-Scale Boosted Dehazing Network with Dense Feature Fusion based on the U-Net architecture. The proposed method is designed two principles, boosting and error feedback, show that they are suitable for dehazing problem. By incorporating Strengthen-Operate-Subtract strategy in decoder of model, develop simple yet effective boosted to progressively restore haze-free image. To address issue preserving spatial information architecture, design dense feature fusion...
Most of the existing deep learning-based dehazing methods are trained and evaluated on image datasets, where dehazed images generated by only exploiting information from corresponding hazy ones. On other hand, video algorithms, which can acquire more satisfying results temporal redundancy neighborhood frames, receive less attention due to absence datasets. Therefore, we propose first REal-world VIdeo DEhazing (REVIDE) dataset be used for supervised learning algorithms. By utilizing a...
In this paper, we propose an end-to-end convolution neural network (CNN) to restore a clear high-resolution image from severely blurry image. It's highly ill-posed problem and brings tremendous challenges state-of-art deblurring or super-resolution (SR) methods. A straightforward way solve is concatenate two types of networks directly. However, experiments show that the concatenation independent increases computation complexity instead generating satisfying images. Consequently, focus on...
Abstract The behavior of bolted connectors with double embedded nuts (BCDENs) in steel‐fiber reinforced concrete (SFRC) remains uncertain, restricting their application steel‐SFRC composite beams. This study explored the shear performance four push‐off test specimens, varying bolt diameters, grades, and strength. analysis covered failure modes, load–slip response, load–strain behavior, resistance, initial slip load, stiffness, peak slip, ductility BCDENs. Bolt shearing off was prevalent mode...
This paper reviews the first NTIRE challenge on video super-resolution (restoration of rich details in low-resolution frames) with focus proposed solutions and results. A new REalistic Diverse Scenes dataset (REDS) was employed. The divided into 2 tracks. Track 1 employed standard bicubic downscaling setup while had realistic dynamic motion blurs. Each competition 124 104 registered participants. There were total 14 teams final testing phase. They gauge state-of-the-art super-resolution.
In this paper, we present two methods for obtaining visual odometry (VO) estimates using a scanning laser rangefinder. Although common VO implementations utilize stereo camera imagery, passive cameras are dependent on ambient light. contrast, actively illuminated sensors such as rangefinders work in variety of lighting conditions, including full darkness. We leverage previous successes by applying sparse appearance‐based to intensity images, and address the issue motion distortion...
Attitude estimation from unknown corresponding points plays a crucial role in visual tasks involving feature points, but current solutions suffer slow computation and poor tolerance to noise. In order address these drawbacks, this article proposes an inertial measurement unit (IMU)-assisted uncertainty-weighted attitude algorithm noncorresponding points. The enhances both speed robustness anisotropic method begins with initial pose generator based on preintegration, optimizing the selection...
Calibration between multiple sensors is a fundamental procedure for data fusion. To address the problems of large errors and tedious operation, we present novel method to conduct calibration light detection ranging (LiDAR) camera. We invent target, which an arbitrary triangular pyramid with three chessboard patterns on its planes. The target contains both 3D information 2D information, can be utilized obtain intrinsic parameters camera extrinsic system. In proposed method, world coordinate...
The emergence of large language models (LLMs) has substantially influenced natural processing, demonstrating exceptional results across various tasks. In this study, we employ ``Introspective Tips" to facilitate LLMs in self-optimizing their decision-making. By introspectively examining trajectories, LLM refines its policy by generating succinct and valuable tips. Our method enhances the agent's performance both few-shot zero-shot learning situations considering three essential scenarios:...
Accurate pose estimation relies on high-quality sensor measurements. Due to manufacturing tolerance, every (camera or lidar) needs be individually calibrated. Feature-based techniques using simple calibration targets (e.g., a checkerboard pattern) have become the dominant approach camera calibration. Existing lidar methods require controlled environment space of known dimension) specific configurations supporting hardware coupled with GPS/IMU). Leveraging recent state developments based...
In this paper, we propose a Multi-Scale Boosted Dehazing Network with Dense Feature Fusion based on the U-Net architecture. The proposed method is designed two principles, boosting and error feedback, show that they are suitable for dehazing problem. By incorporating Strengthen-Operate-Subtract strategy in decoder of model, develop simple yet effective boosted to progressively restore haze-free image. To address issue preserving spatial information architecture, design dense feature fusion...
Pose estimation with unknown correspondences between 3D object points and 2D image is known as the simultaneous pose correspondence determination problem in field of computer vision. It currently still diffcutlt to solve particularly appearance occlusion cluster. In this paper, we present a new iterative algorithm for an without any additional 3D-2D corresondence infromation. Our method combines SoftAssign derterming OI (orthogonal iteration) computing pose. An assignment matrix which...
Single-image super-resolution is a fundamental task for vision applications to enhance the image quality with respect spatial resolution. If input contains degraded pixels, artifacts caused by degradation could be amplified methods. Image blur common source. Images captured moving or still cameras are inevitably affected motion due relative movements between sensors and objects. In this work, we focus on presence of blur. We propose deep gated fusion convolution neural network generate clear...
Image super-resolution is a fundamental pre-processing technique for the machine vision applications of robotics and other mobile platforms. Inevitably, images captured by camera tend to emerge severe motion blur this degradation will deteriorate performance current state-of-the-art methods. In paper, we propose deep dual-branch convolution neural network (CNN) generate clear high-resolution image from single natural with blurs. Compared off-the-shelf methods, our method, called DB-SRN, can...
Due to the limitations of imaging processors and complex weather conditions, image degradation is often inevitable. Existing deep learning-based restoration methods rely on powerful feature representation capacity networks pay less attention inherent properties signal, e.g. variations in spatial scale orientations across image, which makes them ineffective for tasks. In this paper, we propose a Multiscale Gabor Wavelet Network (MsGWN) restoration. We apply multi-scale architecture extract...
Although several algorithms have been presented to solve the simultaneous pose and correspondence estimation problem, correct solution may not be reached with traditional random-start initialization method. In this paper, we derive a novel method which estimates initial value based on genetic algorithm, considering influences of different guesses comprehensively. First, set random is generated as candidate solutions. Second, assignment matrix perspective projection error are computed for...