A5107340706- 3D Shape Modeling and Analysis
- Sparse and Compressive Sensing Techniques
- Remote Sensing and LiDAR Applications
- Photoacoustic and Ultrasonic Imaging
- Image and Signal Denoising Methods
- Astronomical Observations and Instrumentation
- Robotics and Sensor-Based Localization
- Gamma-ray bursts and supernovae
- Human Pose and Action Recognition
- Computer Graphics and Visualization Techniques
- Speech and Audio Processing
- Recycling and Waste Management Techniques
- Optical measurement and interference techniques
- 3D Surveying and Cultural Heritage
- Aerodynamics and Acoustics in Jet Flows
- Advanced Image Processing Techniques
- Advanced Vision and Imaging
- Granular flow and fluidized beds
- Image Processing and 3D Reconstruction
- Underwater Acoustics Research
- Astronomy and Astrophysical Research
- Thermochemical Biomass Conversion Processes
- Generative Adversarial Networks and Image Synthesis
- Image Processing Techniques and Applications
- Astrophysical Phenomena and Observations
Anhui Jianzhu University
2024
National Astronomical Observatories
2022
Jiangsu University of Science and Technology
2022
Nanjing University of Information Science and Technology
2014
Conventional near-field acoustic holography based on compressive sensing either does not fully exploit the underlying block-sparse structures of signal or suffers from a mismatch between actual and predefined block structure due to lack prior information about partitions, resulting in poor accuracy sound field reconstruction. In this paper, pattern-coupled Bayesian method is proposed for sparse reconstruction fields. The establishes hierarchical Gaussian-Gamma probability model with...
In this paper, we propose a novel local descriptor-based framework, called You Only Hypothesize Once (YOHO), for the registration of two unaligned point clouds. contrast to most existing descriptors which rely on fragile reference frame gain rotation invariance, proposed descriptor achieves invariance by recent technologies group equivariant feature learning, brings more robustness density and noise. Meanwhile, in YOHO also has part, enables us estimate from just one correspondence...
With the fast increase in resolution of astronomical images, question how to process and transfer such large images has become a key issue astronomy. We propose new real-time compression reconstruction algorithm for based on compressive sensing techniques. first reconstruct original signal with fewer measurements, according its compressibility. Then, characteristics we apply Daubechies orthogonal wavelets obtain sparse representation. A matrix representing random Fourier ensemble is used...
Matching cross-modality features between images and point clouds is a fundamental problem for image-to-point cloud registration. However, due to the modality difference points, it difficult learn robust discriminative by existing metric learning methods feature matching. Instead of applying on data, we propose unify pretrained large-scale models first, then establish correspondence within same modality. We show that intermediate features, called diffusion extracted depth-to-image are...
We introduce Coverage Axis++, a novel and efficient approach to 3D shape skeletonization. The current state-of-the-art approaches for this task often rely on the watertightness of input or suffer from substantial computational costs, thereby limiting their practicality. To address challenge, Axis++ proposes heuristic algorithm select skeletal points, offering high-accuracy approximation Medial Axis Transform (MAT) while significantly mitigating intensity various representations. simple yet...
We present Surf-D, a novel method for generating high-quality 3D shapes as Surfaces with arbitrary topologies using Diffusion models. Previous methods explored shape generation different representations and they suffer from limited poor geometry details. To generate surfaces of topologies, we use the Unsigned Distance Field (UDF) our surface representation to accommodate topologies. Furthermore, propose new pipeline that employs point-based AutoEncoder learn compact continuous latent space...
We introduce Point2Skeleton, an unsupervised method to learn skeletal representations from point clouds. Existing skeletonization methods are limited tubular shapes and the stringent requirement of watertight input, while our aims produce more generalized for complex structures handle Our key idea is use insights medial axis transform (MAT) capture intrinsic geometric topological natures original input points. first predict a set points by learning transformation, then analyze connectivity...
Lobster eye telescopes are a type of innovative telescope design, which could observe celestial objects over very wide field view in x-ray band. Thanks to this property, lobster widely used detect transients time-domain astronomy. However, images obtained by modified their unique point spread functions, would photons from sources large with crucify structure. Therefore, it is hard design an automatic source detection algorithm high efficiency and fast speed. Manual interventions always...
The first X-ray source catalog of Insight-HXMT Galactic Plane (|b|<10deg) Scanning Survey (GPSS) is presented based on the data accumulated from June 2017 to August 2021. 4 yr limit sensitivities at main energy bands can reach 8.2x10^(-12) erg/s/cm^2} (2-6 keV), 4.21x10^(-11) erg/s/cm^2 (7-40 keV) and 2.78x10^(-11) (25-100 keV). More than 1300 sources have been monitored a wide band (1$-$100\,keV), which 223 signal-to-noise ratio greater 5. We combined GPSS MAXI found it feasible obtain...
In this paper, we present a new method for the multiview registration of point cloud. Previous methods rely on exhaustive pairwise to construct densely-connected pose graph and apply Iteratively Reweighted Least Square (IRLS) compute scan poses. However, constructing is time-consuming contains lots outlier edges, which makes subsequent IRLS struggle find correct To address above problems, first propose use neural network estimate overlap between pairs, enables us sparse but reliable graph....
Sparse representation learning has recently gained a great success in signal and image processing, thanks to recent advances dictionary learning. To this end, the $\ell_0$-norm is often used control sparsity level. Nevertheless, optimization problems based on are non-convex NP-hard. For these reasons, relaxation techniques have been attracting much attention of researchers, by priorly targeting approximation solutions (e.g. $\ell_1$-norm, pursuit strategies). On contrary, paper considers...
This paper deals with sparse coding for dictionary learning in representations. Because involves an ℓ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0</sub> -norm, most, if not all, existing solutions only provide approximate solution. Instead, this paper, a real optimization is considered the problem providing global optimal The proposed method reformulates as Mixed-Integer Quadratic Program (MIQP), allowing then to obtain solution by using...
Abstract：Biomass is gradually perceived as an important carbon neutral and renewable energy source. In general, biomass particles are non-spherical filamentous, their efficiency of use considerably reduced due to the high moisture content. As a result, pretreatment dehydration crucial process in production. The effectiveness removal, however, directly affected by fluidization characteristics chain-like drying facilities. this article, distribution agglomeration fluidized bed were studied...