A5100688517- Cryospheric studies and observations
- Advanced Image and Video Retrieval Techniques
- Remote-Sensing Image Classification
- Winter Sports Injuries and Performance
- Arctic and Antarctic ice dynamics
- Landslides and related hazards
- Geology and Paleoclimatology Research
- Climate change and permafrost
- Advanced Neural Network Applications
- Image Processing Techniques and Applications
- Remote Sensing and Land Use
- Infrared Target Detection Methodologies
- Video Surveillance and Tracking Methods
- Robotics and Sensor-Based Localization
- Geographic Information Systems Studies
- Image Retrieval and Classification Techniques
- Sparse and Compressive Sensing Techniques
- Advanced Measurement and Detection Methods
- Inertial Sensor and Navigation
- Polar Research and Ecology
- Advanced Vision and Imaging
- Image and Object Detection Techniques
- Geophysics and Gravity Measurements
- Climate variability and models
- Image Enhancement Techniques
Beijing Institute of Technology
2013-2025
Northwestern Polytechnical University
2011-2025
Southwest University
2025
Utah State University
2022-2024
Lanzhou University of Technology
2024
Shenyang Aerospace University
2024
Beijing Normal University
2021-2024
Wuhan University
2009-2023
Qufu Normal University
2022
Jilin Agricultural University
2021-2022
Abstract. Ice flow models of the Antarctic ice sheet are commonly used to simulate its future evolution in response different climate scenarios and assess mass loss that would contribute sea level rise. However, there is currently no consensus on estimates balance sheet, primarily because differences representation physical processes, forcings employed initial states models. This study presents results from model simulations 13 international groups focusing during period 2015–2100 as part...
In this paper, a novel deep neural network (DNN)-driven feature learning method is proposed and applied to multi-view facial expression recognition (FER). method, scale invariant transform (SIFT) features corresponding set of landmark points are first extracted from each image. Then, matrix consisting the SIFT vectors used as input data sent well-designed DNN model for optimal discriminative classification. The employs several layers characterize relationship between their high-level...
Antarctica's ice shelves modulate the grounded flow, and weakening of due to climate forcing will decrease their 'buttressing' effect, causing a response in ice. While processes governing ice-shelf are complex, uncertainties sheet also difficult assess. The Antarctic BUttressing Model Intercomparison Project (ABUMIP) compares ice-sheet model responses buttressing by investigating 'end-member' scenario total sustained loss shelves. Although unrealistic, this enables gauging sensitivity an...
Abstract. The sea level contribution of the Antarctic ice sheet constitutes a large uncertainty in future projections. Here we apply linear response theory approach to 16 state-of-the-art models estimate from basal shelf melting within 21st century. purpose this computation is Antarctica's global rise that arises oceanic forcing and associated melting. Ice considered be major if not largest perturbation sheet's flow into ocean. However, by computing only melting, our study neglecting number...
Abstract Projections of the sea level contribution from Greenland and Antarctic ice sheets (GrIS AIS) rely on atmospheric oceanic drivers obtained climate models. The Earth System Models participating in Coupled Model Intercomparison Project phase 6 (CMIP6) generally project greater future warming compared with previous 5 (CMIP5) effort. Here we use four CMIP6 models a selection CMIP5 to force multiple sheet as part Ice Sheet for (ISMIP6). We find that projected at 2100 model ensemble under...
Multi-modal large language models (MLLMs) have demonstrated remarkable success in vision and visual-language tasks within the natural image domain. Owing to significant domain gap between remote sensing (RS) images, development of MLLMs RS is still infant stage. To fill gap, a pioneer MLLM named EarthGPT integrating various multi-sensor interpretation uniformly proposed this paper for universal comprehension. Firstly, visual-enhanced perception mechanism constructed refine incorporate...
In this paper, we establish upper and lower bounds for some statistical estimation problems through concise information-theoretic arguments. Our bound analysis is based on a simple yet general inequality which call the information exponential inequality. We show that naturally leads to randomized method, performance can be obtained. The bounds, applicable all estimators, are obtained by original applications of well known inequalities, approximately match various important problems....
Abstract. Ice sheet numerical modeling is an important tool to estimate the dynamic contribution of Antarctic ice sea level rise over coming centuries. The influence initial conditions on model simulations, however, still unclear. To better understand this influence, state intercomparison exercise (initMIP) has been developed compare, evaluate, and improve initialization procedures their impact century-scale simulations. initMIP first set experiments Sheet Model Intercomparison Project for...
A standard approach in large scale machine learning is distributed stochastic gradient training, which requires the computation of aggregated gradients over multiple nodes on a network. Communication major bottleneck such applications, and recent years, compressed methods as QSGD (quantized SGD) sparse SGD have been proposed to reduce communication. It was also shown that error compensation can be combined with compression achieve better convergence scheme each node compresses its local...
Abstract. We introduce MPAS-Albany Land Ice (MALI) v6.0, a new variable-resolution land ice model that uses unstructured Voronoi grids on plane or sphere. MALI is built using the Model for Prediction Across Scales (MPAS) framework developing Earth system components and Albany multi-physics code base solution of coupled systems partial differential equations, which itself makes use Trilinos solver libraries. includes three-dimensional first-order momentum balance (Blatter–Pattyn) by linking...
In planetary science, it is an important basic work to recognize and classify the features of topography geomorphology from massive data remote sensing. Therefore, this article proposes a lightweight model based on VGG-16, which can selectively extract some sensing images, remove redundant information, images. This not only ensures accuracy, but also reduces parameters model. According our experimental results, has great improvement in image classification, original accuracy 85%-98% now. At...
Object detection is an essential task in computer vision. Recently, several convolution neural network (CNN)-based detectors have achieved a great success natural scenes. However, for optical remote sensing images with large scale of view, lower proportion foreground target pixels and drastic differences object present considerable challenges. To address these problems, we propose novel one-stage detector called the full-scale (FSoD-Net) which consists proposed multiscale enhancement...
Remote sensing image classification (RSIC) is a classical and fundamental task in the intelligent interpretation of remote imagery, which can provide unique labeling information for each acquired image. Thanks to potent global context extraction ability multi-head self-attention (MSA) mechanism, visual transformer (ViT)-based architectures have shown excellent capability natural scene classification. However, order achieve powerful RSIC performance, it insufficient capture spatial alone....
Automatic clothes search in consumer photos is not a trivial problem as are usually taken under completely uncontrolled realistic imaging conditions. In this paper, novel framework presented to tackle issue by leveraging low-level features (e.g., color) and high-level (attributes) of clothes. First, content-based image retrieval(CBIR) approach based on the bag-of-visual-words (BOW) model developed our baseline system, which codebook constructed from extracted dominant color patches. A...
Optical remote sensing object detection is a challenging task, because of the complex background interference, ambiguous appearances tiny objects, densely arranged circumstances, and multiclass with vaster scale variances irregular aspect ratios. The performance seriously restricted. Thus, in this article, inspired by anchor-free framework, aiming to solve these difficulties improve optical performance, powerful one-stage detector multiscale semantic fusion-guided fractal convolution network...
Abstract. The Antarctic Ice Sheet represents the largest source of uncertainty in future sea level rise projections, with a contribution to by 2100 ranging from −5 43 cm equivalent under high carbon emission scenarios estimated recent Model Intercomparison for CMIP6 (ISMIP6). ISMIP6 highlighted different behaviors East and West ice sheets, as well possible role increased surface mass balance offsetting dynamic loss response changing oceanic conditions shelf cavities. However, detailed...
The widely distributed glaciers in the greater Himalayan region have generally experienced rapid shrinkage since 1850s. As invaluable sources of water and because their scarcity, these are extremely important. Beginning twenty-first century, new methods been applied to measure mass budget glaciers. Investigations shown that albedo is an important parameter affects melting glaciers.The surface based on Moderate Resolution Imaging Spectroradiometer (MODIS) data over Hindu Kush, Karakoram...
The rise of crowdsourcing brings new types malpractices in Internet advertising. One can easily hire web workers through malicious platforms to attack other advertisers. Such human generated crowd frauds are hard detect by conventional fraud detection methods. In this paper, we carefully examine the characteristics group behaviors and identify three persistent patterns, which moderateness, synchronicity dispersivity. Then propose an effective method for search engine advertising based on...
Automatic personal clothing retrieval on photo collections, i.e., searching the same clothes worn by person, is not a trivial problem as photos are usually taken under completely uncontrolled realistic imaging conditions. Typically, captured images have large variations due to geometric deformation, occlusion, cluttered background, and photometric variability from illumination viewpoint, which pose significant challenges text-based or reranking-based visual search methods. In this paper,...
Starcraft II (SC2) is widely considered as the most challenging Real Time Strategy (RTS) game. The underlying challenges include a large observation space, huge (continuous and infinite) action partial observations, simultaneous move for all players, long horizon delayed rewards local decisions. To push frontier of AI research, Deepmind Blizzard jointly developed StarCraft Learning Environment (SC2LE) testbench complex decision making systems. SC2LE provides few mini games such MoveToBeacon,...
Currently, under supervised learning, a model pre-trained by large-scale nature scene dataset and then fine-tuned on few specific task labeling data is the paradigm that has dominated knowledge transfer learning. Unfortunately, due to different categories of imaging stiff challenges annotation, there not large enough uniform remote sensing support pre-training in domain (RSD). Moreover, models datasets learning directly fine-tuning diverse downstream tasks seems be crude method, which easily...
Arbitrary-oriented object detection (AOOD) from optical remote sensing imagery has to correctly generate delicate oriented boundary boxes (OBBs) and meanwhile identify their specific categories. However, how make detectors learn parameters of OBBs, especially for the crucial orientation information, category complex background becomes a challenge task. Therefore, in this article, exploring better way guide detector parametric information novel one-stage anchor-free called Posterior Instance...