A5057164374- Advanced Image and Video Retrieval Techniques
- Advanced Neural Network Applications
- Robotics and Sensor-Based Localization
- Coastal wetland ecosystem dynamics
- Remote Sensing and LiDAR Applications
- Aquatic Ecosystems and Phytoplankton Dynamics
- Remote-Sensing Image Classification
- Visual Attention and Saliency Detection
- Ecology and Vegetation Dynamics Studies
- Video Surveillance and Tracking Methods
- Domain Adaptation and Few-Shot Learning
- Automated Road and Building Extraction
- Plant responses to water stress
- Remote Sensing and Land Use
- Robotic Path Planning Algorithms
- Infrared Target Detection Methodologies
- Biocrusts and Microbial Ecology
- Environmental Changes in China
- Soil Carbon and Nitrogen Dynamics
- Image and Object Detection Techniques
- Regional Development and Environment
- Optical measurement and interference techniques
- Advanced Vision and Imaging
- Multimodal Machine Learning Applications
- GABA and Rice Research
Wuhan University
2007-2024
ZheJiang Academy of Agricultural Sciences
2021-2024
Nanjing University of Information Science and Technology
2023
China Meteorological Administration
2023
Huawei Technologies (China)
2022
Group Sense (China)
2022
Chinese Academy of Medical Sciences & Peking Union Medical College
2018
Wenzhou Polytechnic
2013
Zhejiang Institute of Subtropical Crops
2011
Research Institute of Subtropical Foresty
2011
One fundamental problem in Earth Vision is to accurately find the locations and identify categories of interesting objects aerial images, for which oriented bounding boxes (OBBs) are usually employed depict better emerging with arbitrary orientations. However, regression OBBs always suffers from ambiguous definition targets, often reduces convergency efficiency decreases detection accuracy. Although there some methods like binary segmentation map that can handle this problem, it brings a new...
Detecting tiny objects is a very challenging problem since object only contains few pixels in size. We demonstrate that state-of-the-art detectors do not produce satisfactory results on due to the lack of appearance information. Our key observation Intersection over Union (IoU) based metrics such as IoU itself and its extensions are sensitive location deviation objects, drastically deteriorate detection performance when used anchor-based detectors. To alleviate this, we propose new...
Object detection in Earth Vision has achieved great progress recent years. However, tiny object aerial images remains a very challenging problem since the objects contain small number of pixels and are easily confused with background. To advance research images, we present new dataset for Tiny Detection Aerial Images (AI-TOD). Specifically, AI-TOD comes 700,621 instances eight categories across 28,036 images. Compared to existing datasets mean size is about 12.8 pixels, which much smaller...
Object detection in aerial images is a fundamental yet challenging task remote sensing field. As most objects are arbitrary orientations, oriented bounding boxes (OBBs) have great superiority compared with traditional horizontal (HBBs). However, the regression-based OBB methods always suffer from ambiguity definition of learning targets, which will decrease accuracy. In this paper, we provide comprehensive analysis representations and cast regression as pixel-level classification problem,...
Detecting arbitrarily oriented tiny objects poses intense challenges to existing detectors, especially for label assignment. Despite the exploration of adaptive assignment in recent object extreme geometry shape and limited feature still induce severe mismatch imbalance issues. Specifically, position prior, positive sample feature, instance are mismatched, learning extreme-shaped is biased unbalanced due little proper supervision. To tackle these issues, we propose a dynamic prior along with...
Object detection has achieved great progress with the development of anchor-based and anchor-free detectors. However, tiny objects is still challenging due to lack appearance information. In this paper, we observe that Intersection over Union (IoU), most widely used metric in object detection, sensitive slight offsets between predicted bounding boxes ground truths when detecting objects. Although some new metrics such as GIoU, DIoU CIoU are proposed, their performance on below expected level...
The Vision Meets Drone (VisDrone2019) Single Object Tracking challenge is the second annual research activity focusing on evaluating single-object tracking algorithms drones, held in conjunction with International Conference Computer (ICCV 2019). VisDrone-SOT2019 Challenge goes beyond its VisDrone-SOT2018 predecessor by introducing 25 more challenging sequences for long-term tracking. We evaluate and discuss results of 22 participating 19 state-of-the-art trackers collected dataset. are...
Extracting building footprints from aerial images is essential for precise urban mapping with photogrammetric computer vision technologies. Existing approaches mainly assume that the roof and footprint of a are well overlapped, which may not hold in off-nadir as there often big offset between them. In this paper, we propose an vector learning scheme, turns extraction problem into instance-level joint prediction its corresponding "roof to footprint" vector. Thus can be estimated by...
Mangrove forests are colloquially referred to as “Earth’s kidneys” and serve many important ecological commercial functions. Salinity waterlogging stress the most abiotic stressors restricting growth development of mangroves. Kandelia obovata (K. obovata) is greatest latitudinally-distributed salt mangrove species in China.Here, morphology transcriptomics were used study response K. stress. In addition, weighted gene co-expression network analysis combined expression phenotypic datasets was...
Summary 1. To test whether clonal macrophytes can select favourable habitats in heterogeneous environments, fragments of the stoloniferous submerged macrophyte Vallisneria spiralis were subjected to conditions which light intensity and substratum nutrients patchily distributed. The allocation biomass accumulation ramet production clones different patches was examined. 2. proportion both number allocated rich significantly higher than poor patches. greatest values clone leaf produced...
In this paper, we propose a new dataset and benchmark for low altitude UAV object detection, aiming to find localize waste plastic bottles in the wild, as well inspire development of detection models be capable detecting small transparent objects. To end, collect 25, 407 images with various kinds backgrounds. Unlike traditional horizontal bounding box based annotation methods, use oriented accurately compactly annotate bottles, which provides more detailed information subsequent robotic...
3D building reconstruction from monocular remote sensing imagery is an important research problem and economic solution to large-scale city modeling, compared with LiDAR data multi-view imagery. However, several challenges such as the partial invisibility of footprints facades, serious shadow effect, extreme variance height in areas, have restricted existing image based studies certain application scenes, i.e., modeling simple low-rise buildings near-nadir images. In this study, we propose a...
The scientific management of salinized agricultural lands and the use undeveloped saline to ensure food security have become one most urgent tasks nowadays. Biochar contains rich carbon (C) functional groups, processes high alkalinity, porosity, specific surface area (SSA). Thus, it has been widely used as an effective organic conditioner in acidic or neutral soils improve their fertility. However, so far, impacts biochar application on properities underlying mechanisms remain unveiled....
Unmanned aerial vehicle (UAV)-based imaging systems have many superiorities compared with other platforms, such as high flexibility and low cost in collecting images, providing wide application prospects. However, the acquisition of UAV-based image commonly results very resolution large-scale which poses great challenges for subsequent applications. Therefore, an efficient representation UAV images is necessary extraction required information a reasonable time. In this work, we proposed...
We conducted a single-arm prospective phase II study to determine the efficacy and safety of first-line treatment advanced pancreatic cancer with nab-paclitaxel S-1 followed by maintenance therapy. Nab-paclitaxel was administered intravenously on days 1 8 at 120 mg/m2. mg/day (for surface area ≥ 1.5 m2), 100 between 1.25–1.5 80 < 1.25 m2) were given two times daily 1–14 every 3 weeks. Patients who achieved response stable disease after 6 cycles in same schedule until progression or...
Wind turbine detecting and blade tip positioning are essential parts of UAV-based automatic inspection. Previous methods which adopt LiDAR or ultrasonic sensor may cause cost increasement system instability. We propose a vision-based approach for positioning. More precisely, we first detect each structure wind by combining Mask R-CNN detector shape constraints. The pixel coordinates tips can be extracted then used to solve PnP problem. Finally get GPS useful achieving path planning during...
3D building reconstruction from monocular remote sensing imagery is an important research problem that has been extensively studied for several decades. Although a more economic data source compared with the LiDAR and multi-view imagery, its limited information results in great challenges restricts performance of existing methods. Moreover, expensive cost quantity annotations also restrict application scenes methods, which are mostly based on fully-supervised learning. In our previous work,...
Forest fires are a major natural calamity that inflict substantial harm on forest resources and the socio-economic landscape. The eastern region of China is particularly susceptible to frequent fires, characterized by high population density vibrant economic activities. Precise forecasting in this area essential for devising effective prevention strategies. This research utilizes blend kernel analysis, autocorrelation standard deviation ellipse method, augmented geographic information...