A5040395839- Advanced Image and Video Retrieval Techniques
- Robotics and Sensor-Based Localization
- Video Surveillance and Tracking Methods
- Indoor and Outdoor Localization Technologies
- Advanced Measurement and Detection Methods
- Human Pose and Action Recognition
- Mobile Ad Hoc Networks
- Advanced Algorithms and Applications
- Advanced Vision and Imaging
- Image Retrieval and Classification Techniques
- Advanced Computational Techniques and Applications
- Rough Sets and Fuzzy Logic
- Anomaly Detection Techniques and Applications
- Advanced Image Processing Techniques
- Industrial Technology and Control Systems
- Ultra-Wideband Communications Technology
- Robotic Path Planning Algorithms
- Advanced Neural Network Applications
- Video Analysis and Summarization
- Opportunistic and Delay-Tolerant Networks
- Data Mining Algorithms and Applications
- Target Tracking and Data Fusion in Sensor Networks
- Image Processing Techniques and Applications
- Data Management and Algorithms
- Vehicle Dynamics and Control Systems
Nanjing University of Information Science and Technology
2021-2025
Southeast University
2016-2024
Nanjing University
2015-2024
Shenzhen University
2023
Beijing Forestry University
2023
Beijing University of Posts and Telecommunications
2009-2022
Jilin University
2018-2022
University of Jinan
2004-2021
KDDI Research (Japan)
2021
Hubei University of Automotive Technology
2019
We introduce an online learning approach for multitarget tracking. Detection responses are gradually associated into tracklets in multiple levels to produce final tracks. Unlike most previous approaches which only focus on producing discriminative motion and appearance models all targets, we further consider features distinguishing difficult pairs of targets. The tracking problem is formulated using learned CRF model, transformed energy minimization problem. functions include a set unary...
We describe an online approach to learn non-linear motion patterns and robust appearance models for multi-target tracking in a tracklet association framework. Unlike most previous approaches that use linear methods only, we build map better explain direction changes produce more affinities between tracklets. Moreover, based on the incremental learned entry/exit map, multiple instance learning method is devised strong tracking; positive sample pairs are collected from different track-lets so...
Colors with high saturation are of prime significance for display and imaging devices. So far, structural colors arising from all-dielectric metasurfaces, particularly amorphous silicon titanium oxide, have exceeded the gamut standard RGB (sRGB) space. However, excitation higher-order modes dielectric materials hinders further increase saturation. Here, to address challenge, we propose a new design strategy multipolar-modulated metasurfaces multi-dielectric stacked layers realize deep...
We propose a learning-based Conditional Random Field (CRF) model for tracking multiple targets by progressively associating detection responses into long tracks. Tracking task is transformed data association problem, and most previous approaches developed heuristical parametric models or learning evaluating independent affinities between track fragments (tracklets). argue that the assumption not valid in many cases, adopt CRF to consider both tracklet dependencies among them, which are...
Ultra-wide band (UWB) localization system suffers from deteriorating performance in complex scenes, especially non-line-of-sight (NLOS) conditions. In order to improve the accuracy and robustness of NLOS environments, we propose an end-to-end deep neural network with both distance received signal strength (RSS) measurements. On one hand, high-level spatial-temporal features can be learned through proposed RSS data, which benefits performance. other is robust variance number available...
We have developed a facile and compatible method to in situ fabricate uniform metal nanowire networks on substrates. The as-fabricated show low sheet resistance high transmittance (2.2 Ω sq–1 at T = 91.1%), which is equivalent that of the state-of-the-art networks. demonstrated becomes homogeneous from deep-ultraviolet (200 nm) near-infrared (2000 when size wire spacing increases micrometer size. Theoretical experimental analyses indicated we can improve conductivity as well keep their by...
In this article, a resilient tightly coupled ultra-wideband (UWB) visual–inertial indoor localization system (R-UVIS) is developed to obtain accurate and robust performance in complex scenes, even the case when sensors fail. More specifically, three schemes are designed for proposed system. First, we introduce line image patch features improve precision robustness of visual features. Besides, propose loop closure relocalization methods based on multifeatures Second, UWB sensor into suppress...
In this article, we propose a self-supervised deep network that mainly applies location and ranging error corrections to classical ultrawideband (UWB) localization approach improve its accuracy. The core of method is the learning strategy which removes requirement for ground truth in training process thus reduces cost method. To end, first use an existing UWB provide initial tag location, then build correction (DLRC) jointly estimate position distance corrections. built based on these...
SIFT (scale invariant feature transform) is an important local descriptor. Since its expensive computation, SURF (speeded-up robust features) proposed. Both of them are designed mainly for gray images. However, color provides valuable information in object description and matching tasks. To overcome the drawback to increase descriptor's distinctiveness, this paper presents a novel descriptor which combines kernel histograms Haar wavelet responses construct vector. So two elements In image...
Indoor positioning without GPS is a challenge task, especially, in complex scenes or when sensors fail. In this paper, we develop an ultra-wideband aided visual-inertial system (UVIP) which aims to achieve accurate and robust results indoor environments. To end, point-line-based stereo odometry (PL-sVIO) firstly designed improve the accuracy structured low-textured scenarios by making use of line features. Secondly, loop closure method proposed suppress drift PL-sVIO based on image patch...
SiO2/Si core-shell nanowires array coated with gap-rich silver nanoparticles were demonstrated as a highly reproducible surface-enhanced Raman scattering (SERS) substrate. SERS detection of relative standard deviation 8% for 10−4 M R6G spot size ∼2 μm and 900 spots over an area 150 × μm2 was reported. The high reproducibility is ascribed to the polarization-independent electrical field distribution among three-dimensional nanowire structure optimized thickness SiO2 shell layer.
Disorder is often considered the opposite of order, lacking quantitative methods and being difficult to control. Disordered nanostructures can be conveniently prepared by bottom-up approaches, such as self-assembly, but their intrinsic randomness lead unpredictable results, impeding reproducibility application. Here, we demonstrate that deterministic, angle-dependent visual appearances induced specific correlated disorder achieved through reveal plenty room for tailoring color appearance...
VANET (Vehicular ad hoc Network) is a special kind of wireless network where every single node vehicle moving in relatively high velocity, which leads to exclusive challenges like rapid changing topologies, safety and privacy concerns. In this specific network, broadcasts tend be carrying important messages such as car accident notification, disaster alert or extreme traffic condition. Thus the propagation broadcasted emergency could critical save human lives property. Many researchers have...
Learning-based TOA-UWB localization methods have been developed rapidly in recent years and achieve state-of-the-art results complex scenes. However, they still suffer from two drawbacks: 1) biased measurements with large noise are not suppressed effectively, 2) geometric information which is important for UWB considered. Thus, we propose twofold strategies this paper to overcome these issues: A novel deep attention-based network proposed. In network, introduce the transformer encoder learn...
This paper presents a novel multiple objects tracking method by constructing an improved tracklet affinity function to enhance the performance of object (MOT) within network flow optimization framework. The aims this are improve appearance model and motion function, both models being key factors limit MOT in detection In proposed method, association is considered as generalized linear assignment problem relied on tracklets that calculated sparse representation rank-based model. uses...