- Advanced Image and Video Retrieval Techniques
- Image Processing Techniques and Applications
- Digital Media Forensic Detection
- Handwritten Text Recognition Techniques
- Image Retrieval and Classification Techniques
- Advanced Neural Network Applications
- Simulation and Modeling Applications
- Semantic Web and Ontologies
- Advanced Image Processing Techniques
- Advanced Vision and Imaging
- Optical Network Technologies
- Advanced Photonic Communication Systems
- Robotics and Sensor-Based Localization
- Advanced Optical Network Technologies
- Natural Language Processing Techniques
- Image Processing and 3D Reconstruction
- Video Surveillance and Tracking Methods
- Medical Image Segmentation Techniques
- Advanced Computational Techniques and Applications
- Remote Sensing and Land Use
- Advanced Steganography and Watermarking Techniques
- Face and Expression Recognition
- Advanced Algorithms and Applications
- Image and Object Detection Techniques
- Advanced Image Fusion Techniques
Beihang University
2009-2025
Institute of Forensic Science
2009-2025
University of Science and Technology of China
2005-2024
Jimei University
2023-2024
Shanghai Zhangjiang Laboratory
2023-2024
University of Electronic Science and Technology of China
2020-2024
Northwest A&F University
2023
Guangxi University of Science and Technology
2022-2023
State Key Laboratory of Software Development Environment
2023
Beijing University of Technology
2010-2023
The detection of fastener defects is an important task in railway inspection systems, and it frequently performed to ensure the safety train traffic. Traditional usually operated by trained workers who walk along lines search for potential risks. However, manual very slow, costly, dangerous. This paper proposes automatic visual system detecting partially worn completely missing fasteners using probabilistic topic model. Specifically, our method able simultaneously model diverse types with...
The fruit quality and yield of sweet peppers can be effectively improved by accurately efficiently controlling the growth conditions taking timely corresponding measures to manage planting process dynamically. use deep-learning-based image recognition technology segment pepper instances is an important means achieving above goals. However, accuracy existing instance segmentation algorithms seriously affected complex scenes such as changes in ambient light shade, similarity between color...
Spatiotemporal data mining plays an important role in air quality monitoring, crowd flow modeling, and climate forecasting. However, the originally collected spatiotemporal real-world scenarios is usually incomplete due to sensor failures or transmission loss. imputation aims fill missing values according observed underlying dependence of them. The previous dominant models impute autoregressively suffer from problem error accumulation. As emerging powerful generative models, diffusion...
The detection of fine-grained objects in remote sensing images has been recognized as a challenging issue, which cannot be well addressed by the existing deep learning-based methods due to their inadaptability multi-scale objects, slow convergence speed, and limitations scarce datasets. To cope with above we propose convolutional neural network (CNN)-based method for complex scenarios. Specifically, adopt CNN residual structure backbone network, extract deep-level details from image....
Subcellular location of a protein is one the key functional characters as proteins must be localized correctly at subcellular level to have normal biological function. In this paper, novel method named LOCSVMPSI has been introduced, which based on support vector machine (SVM) and position-specific scoring matrix generated from profiles PSI-BLAST. With jackknife test RH2427 data set, achieved high overall prediction accuracy 90.2%, higher than results by SubLoc ESLpred set. addition,...
This paper has proposed an architecture of optimised SIFT (scale invariant feature transform) detection for FPGA implementation image matcher. In order based matcher to be implemented on efficiently, in terms speed and hardware resource usage, the original algorithm been significantly following aspects: 1) upsampling replaced with downsampling save interpolation operation. 2) Only four scales two octaves are needed our moderate degradation matching performance. 3) The total dimension...
In this work, we propose a new framework, called Document Image Transformer (DocTr), to address the issue of geometry and illumination distortion document images. Specifically, DocTr consists geometric unwarping transformer an correction transformer. By setting set learned query embedding, captures global context image by self-attention mechanism decodes pixel-wise displacement solution correct distortion. After unwarping, our further removes shading artifacts improve visual quality OCR...
The automatic generation of a text summary is task generating short for relatively long document by capturing its key information. In the past, supervised statistical machine learning was widely used this Automatic Text Summarization (ATS) task, but due to high dependence on quality features, generated summaries lack accuracy and coherence, while computational power involved, performance achieved, could not easily meet current needs. This paper proposes four novel ATS models with...
Aiming at the water hammer issues of hydraulic power system wide-body aircraft with 5000 psi pressure system, based on introduction to theory and key components, a complete simulation model is constructed using AMESim software, simulates flow characteristics under transient large conditions. Through verification, current configuration meets standards maximum condition, but does not meet requirements steady state condition. Further improvements can be made by optimizing response speed servo...
Aims: To enhance the safety of intelligent vehicle drivers and help them achieve active collision avoidance. Background: Existing research on vehicles’ avoidance focuses reference path optimization stability tracking control. However, there is relatively little in emergency accident scenarios. Longitudinal braking technology has been mature, while steering still needs a long time to develop. The joint situations limited. Objective: Modeling longitudinal lateral distances based conditions...
In recent years, tremendous efforts have been made on document image rectification, but existing advanced algorithms are limited to processing restricted images, i.e., the input images must incorporate a complete document. Once captured merely involves local text region, its rectification quality is degraded and unsatisfactory. Our previously proposed DocTr, transformer-assisted network for also suffers from this limitation. work, we present DocTr++, novel unified framework without any...
In fisheye images, rich distinct distortion patterns are regularly distributed in the image plane. These independent of visual content and provide informative cues for rectification. To make best such rectification cues, we introduce SimFIR, a simple framework based on self-supervised representation learning. Technically, first split into multiple patches extract their representations with Vision Transformer (ViT). learn fine-grained representations, then associate different specific model,...
Contour-based instance segmentation has been actively studied, thanks to its flexibility and elegance in processing visual objects within complex backgrounds. In this work, we propose a novel deep network architecture, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i.e</i> ., PolySnake, for generic contour-based segmentation. Motivated by the classic Snake algorithm, proposed PolySnake achieves superior robust performance with an iterative...
Orchard spraying robots must visually obtain citrus tree crown growth information to meet the variable growth-stage-based requirements. However, complex environments and characteristics of fruit trees affect accuracy segmentation. Therefore, we propose a feature-map-based squeeze-and-excitation UNet++ (MSEU) region-based convolutional neural network (R-CNN) segmentation method that intakes red–green–blue-depth (RGB-D) images are pixel aligned visual distance-adjusted eliminate noise. Our...
This paper uses the opening speech given by president Obama at a prime time news conference commemorating his first 100 th day in office as sample and tries to explore how interpersonal meaning is achieved from perspective of Functional Grammar with focus on mood, modal auxiliary, personal pronouns pronoun system, tense shift.The finding shows that makes full use language achieve political purpose using different devices fulfill meaning.
This paper presents IP-SLT, a simple yet effective framework for sign language translation (SLT). Our IP-SLT adopts recurrent structure and enhances the semantic representation (prototype) of input video via an iterative refinement manner. idea mimics behavior human reading, where sentence can be digested repeatedly, till reaching accurate understanding. Technically, consists feature extraction, prototype initialization, refinement. The initialization module generates initial based on visual...
Intensity-based fringe projection profilometry (IBFPP) is used widely because of its simple structure, high robustness, and noise resilience. Most IBFPP methods assume that any scene point illuminated by direct illumination only, but global effects introduce strong biases in the reconstruction result for many real-world scenes. To solve this problem, paper describes an efficient method reconstructing three-dimensional geometry presence illumination. First, average intensity two sinusoidal...
Cost effectiveness and energy efficiency have recently become two practical issues during access networks construction operation. In this paper, we extend our previous proposal of a remote channel combine/split module for long-reach wavelength division multiplexed time passive optical network systems. Besides previously discussed benefits on investment cost under take-up-rate-adaptive mode, study shows that proposed systems help achieve higher saving by adopting traffic-adaptive power...
This paper explores how the logical difference between two ontologies can be tracked using a forgetting-based or uniform interpolation (UI)-based approach. The idea is that rather than computing all entailments of one ontology not entailed by other ontology, which would computationally infeasible, only strongest in are computed. To overcome drawbacks existing forgetting/uniform tools we introduce new forgetting method designed for task different versions large-scale ontologies. sound and...
By comparing and analyzing several common indoor positioning system solutions, we found that System based on wireless LAN is superior, in which fingerprint matching algorithm was widely used had a large room to be improved for more precise performance. This paper presents an algorithm, reduces interference by Gaussian filter, calculates distance accurately weighted centroid algorithm. Furthermore, the improves precision local prediction model revised entire increase accuracy of positioning....