- Advanced Vision and Imaging
- Optical measurement and interference techniques
- Image and Object Detection Techniques
- Robotics and Sensor-Based Localization
- Image Processing Techniques and Applications
- Advanced Image and Video Retrieval Techniques
- 3D Surveying and Cultural Heritage
- Domain Adaptation and Few-Shot Learning
- Advanced Graph Neural Networks
- Topic Modeling
- Time Series Analysis and Forecasting
- Face and Expression Recognition
- 3D Shape Modeling and Analysis
- Advanced Numerical Analysis Techniques
- Biometric Identification and Security
- Face recognition and analysis
- Music and Audio Processing
- Visual Attention and Saliency Detection
- Multimodal Machine Learning Applications
- Vehicle License Plate Recognition
- Remote Sensing and LiDAR Applications
- Anomaly Detection Techniques and Applications
- Seismic Imaging and Inversion Techniques
- Video Surveillance and Tracking Methods
- Fault Detection and Control Systems
Peking University
2015-2024
Dongyang People's Hospital
2019
Wenzhou Medical University
2019
Chinese Academy of Sciences
2003-2004
Institute of Automation
2004
Recent years have witnessed a growing interest in co-object segmentation and multi-modal salient object detection. Many efforts are devoted to segmenting co-existed objects among group of images or detecting from different modalities. Albeit the appreciable performance achieved on respective benchmarks, each these methods is limited specific task cannot be generalized other tasks. In this paper, we develop <bold xmlns:mml="http://www.w3.org/1998/Math/MathML"...
Central catadioptric cameras are imaging devices that use mirrors to enhance the field of view while preserving a single effective viewpoint. In this paper, we propose novel method for calibration central using geometric invariants. Lines and spheres in space all projected into conics image plane. We prove projection line can provide three invariants whereas sphere only two. From these invariants, constraint equations intrinsic parameters camera derived. Therefore, there two kinds variants...
Neural Radiance Fields (NeRF) often struggle with reconstructing and rendering highly reflective scenes. Recent advancements have developed various reflection-aware appearance models to enhance NeRF's capability render specular reflections. However, the robust reconstruction of scenes is still hindered by inherent shape ambiguity on surfaces. Existing methods typically rely additional geometry priors regularize prediction, but this can lead oversmoothed in complex Observing critical role...
This paper presents \textbf{FreEformer}, a simple yet effective model that leverages \textbf{Fre}quency \textbf{E}nhanced Trans\textbf{former} for multivariate time series forecasting. Our work is based on the assumption frequency spectrum provides global perspective composition of across various frequencies and highly suitable robust representation learning. Specifically, we first convert into complex domain using Discrete Fourier Transform (DFT). The Transformer architecture then applied...
Neural Radiance Fields (NeRF) often struggle with reconstructing and rendering highly reflective scenes. Recent advancements have developed various reflection-aware appearance models to enhance NeRF's capability render specular reflections. However, the robust reconstruction of scenes is still hindered by inherent shape ambiguity on surfaces. Existing methods typically rely additional geometry priors regularize prediction, but this can lead oversmoothed in complex Observing critical role...
A spherical object has been introduced into camera calibration for several years through utilizing the properties of an image conic, which is projection occluding contour a sphere in perspective image. However, literature, only algebraic interpretation was presented relation between absolute conic and images. In this paper, we propose two geometric interpretations novel methods using images are derived from these interpretations.
Camera calibration directly from image sequences of a pedestrian without using any object is really challenging task and should be well solved in computer vision, especially visual surveillance. In this paper, we propose novel camera method based on recovering the three orthogonal vanishing points (TOVPs), just an sequence walking straight line, assumption scenes or motions, e.g., control with known 3D coordinates, parallel perpendicular lines, non-natural pre-designed special human as often...
Previous works on vanishing point detection usually use geometric prior for line segment clustering. We find that image context can also contribute to accurate classification. Based this observation, we propose classify segments into three groups according unknown-but-sought points with Manhattan world assumption, using both information and in work. To achieve goal, a novel Transformer based Line Classifier (TLC) group images estimate the corresponding points. In TLC, design descriptor...
If an object is interreflected between two planar mirrors, we may take image containing both the and its multiple reflections, i.e., simultaneously imaging views of by a single pinhole camera. This paper emphasizes problem recovering intrinsic extrinsic parameters camera using silhouettes from one image. View pairs among in can be divided into kinds relationship pair: reflected some mirror (real or virtual) circular motion. Epipoles first kind easily determined intersections common tangent...
The effects of radial lens distortion often appear in wide-angle cameras surveillance and safeguard systems, which may severely degrade performances previous face recognition algorithms. Traditional methods for correction usually employ line features scenarios that are not suitable images. In this paper, we propose a distortion-invariant system called RDCFace, directly only utilize the distorted images faces, to alleviate distortion. RDCFace is an end-to-end trainable cascade network, can...
A line in space is projected to a conic the central catadioptric image, and such called image. This work proposes novel approach for efficiently detecting images using Hough transform. Detecting brings two challenges detection: one that effects of occlusion are very significant where traditional methods may fail, other can belong any type conic, as, line, circle, ellipse, hyperbola, parabola etc., it difficult detect when its unknown. The main contribution this we prove image be...
In Hartley-Kang's paper [7], they directly treated a planar calibration pattern as an image to construct pair together with radial distorted of the pattern, and then proposed very efficient method determine center distortion by estimating epipole in image. After determined distortion, least square was utilized recover function using monotonicity constraints. this paper, we present convex optimization same constraints those required method, whereas our can obtain better results correction....
Radial lens distortion is the most significant in current cameras, and many models are proposed to describe it. Fitzgibbon presented prestigious division model with just a single parameter. Based on fact that line 3D space may be projected onto curve image plane due radial distortion, Alemán et al. utilized Hough transform three parameters automatically correct namely, one parameter coming from model, other two arising corrected line. However, some cases, especially for wide angle lenses,...
Domain adaptation for 3D point cloud has attracted a lot of interest since it can avoid the time-consuming labeling process data to some extent. A recent work named xMUDA leveraged multi-modal domain task semantic segmentation by mimicking predictions between 2D and modalities, outperformed previous single modality methods only using clouds. Based on it, in this paper, we propose novel cross-modal contrastive learning scheme further improve effects. By employing constraints from...
Concentric circles are often used as calibration features since they possess good geometric properties. This paper presents an efficient method for the detection of projected concentric in image plane while considering their special The proposed is capable detecting partially visible circles. Experimental results demonstrate validity approach.