A5101465277- Image and Signal Denoising Methods
- Advanced Image Fusion Techniques
- Remote Sensing and LiDAR Applications
- Advanced Image Processing Techniques
- Remote Sensing in Agriculture
- Face and Expression Recognition
- Smart Agriculture and AI
- Image Enhancement Techniques
- Traumatic Brain Injury and Neurovascular Disturbances
- Advanced Algorithms and Applications
- Forest ecology and management
- Remote Sensing and Land Use
- Forest Ecology and Biodiversity Studies
- Image Processing Techniques and Applications
- Remote-Sensing Image Classification
- Traumatic Brain Injury Research
- Natural Language Processing Techniques
- Video Surveillance and Tracking Methods
- Advanced Decision-Making Techniques
- Plant-Microbe Interactions and Immunity
- Guidance and Control Systems
- Spacecraft Dynamics and Control
- Fungal Plant Pathogen Control
- Topic Modeling
- Geoscience and Mining Technology
Xidian University
2009-2025
Hainan University
2024
Hebei University of Engineering
2024
Northeast Agricultural University
2023
Chinese Academy of Sciences
2015-2021
Institute of Botany
2015-2021
University of Chinese Academy of Sciences
2016-2020
Nanjing University
2019
Czech Academy of Sciences, Institute of Botany
2016-2018
Shanghai Jiao Tong University
2000-2014
Deep neural networks (DNNs) have shown very promising results for various image restoration (IR) tasks. However, the design of network architectures remains a major challenging achieving further improvements. While most existing DNN-based methods solve IR problems by directly mapping low quality images to desirable high-quality images, observation models characterizing degradation processes been largely ignored. In this paper, we first propose denoising-based algorithm, whose iterative steps...
The trade-off between spatial and spectral resolution is one of the fundamental issues in hyperspectral images (HSI). Given challenges directly acquiring high-resolution (HR-HSI), a compromised solution to fuse pair images: has (HR) domain but low-resolution (LR) spectral-domain other vice versa. Model-based image fusion methods including pan-sharpening aim at reconstructing HR-HSI by solving manually designed objective functions. However, such hand-crafted prior often leads inevitable...
In recent decades, global biodiversity has gradually diminished due to the increasing pressure from anthropogenic activities and climatic change. Accurate estimations of spatially continuous three-dimensional (3D) vegetation structures terrain information are prerequisites for studies, which usually unavailable in current ecosystem-wide studies. Although airborne lidar technique been successfully used mapping 3D at landscape regional scales, relatively high cost flight mission significantly...
The rapid development of light detection and ranging (Lidar) provides a promising way to obtain three-dimensional (3D) phenotype traits with its high ability recording accurate 3D laser points. Recently, Lidar has been widely used data in the greenhouse field along other sensors. Individual maize segmentation is prerequisite for throughput extraction at individual crop or leaf level, which still huge challenge. Deep learning, state-of-the-art machine learning method, shown performance object...
Accurate and high throughput extraction of crop phenotypic traits, as a crucial step molecular breeding, is great importance for yield increasing. However, automatic stem-leaf segmentation prerequisite many precise trait extractions still big challenge. Current works focus on the study 2-D image-based segmentation, which are sensitive to illumination occlusion. Light detection ranging (LiDAR) can obtain accurate 3-D information with its active laser scanning strong penetration ability,...
Maize (Zea mays L.) is the third most consumed grain in world and improving maize yield of great importance food security, especially under global climate change more frequent severe droughts. Due to limitation phenotyping methods, current studies only focused on responses phenotypes certain key growth stages. Although light detection ranging (lidar) technology showed potential acquiring three-dimensional (3D) vegetation information, it has been rarely used monitoring phenotype dynamics at...
Deep learning has found successful applications in restoration of two-dimensional (2-D) images including denoising, dehazing, and superresolution. However, existing deep convolutional neural network (DCNN) architecture cannot fully exploit spatial-spectral correlations three-dimensional (3-D) hyperspectral (HSIs) (directly extending 2-D DCNN into 3-D will significantly increase computational complexity); meantime, unlike images, there is an obstacle caused by the shortage training data for...
Abstract Terrestrial light detection and ranging (lidar) can be used to record the three-dimensional structures of trees. Wood-leaf separation, which aims classify lidar points into wood leaf components, is an essential prerequisite for deriving individual tree characteristics. Previous research has tended use intensity (including a multi-wavelength approach) waveform information wood-leaf but most fundamental from point cloud, i.e., x-, y-, z- coordinates each point, this purpose been...
Separating structural components is important but also challenging for plant phenotyping and precision agriculture. Light detection ranging (LiDAR) technology can potentially overcome these difficulties by providing high quality data. However, there are in automatically classifying segmenting of interest. Deep learning extract complex features, it mostly used with images. Here, we propose a voxel-based convolutional neural network (VCNN) maize stem leaf classification segmentation. Maize...
Many deep learning-based solutions to blind image deblurring estimate the blur representation and reconstruct target from its blurry observation. However, these methods suffer severe performance degradation in real-world scenarios because they ignore important prior information about motion (e.g., is diverse spatially varying). Some have attempted explicitly non-uniform kernels by CNNs, but accurate estimation still challenging due lack of ground truth varying images. To address issues, we...
Abstract Background Precision agriculture is an emerging research field that relies on monitoring and managing variability in phenotypic traits. An important trait biomass, a comprehensive indicator can reflect crop yields. However, non-destructive biomass estimation at fine levels unknown challenging due to the lack of accurate high-throughput data algorithms. Results In this study, we evaluated capability terrestrial light detection ranging (lidar) estimating maize plot, individual plant,...
High-throughput maize phenotyping at both organ and plant levels plays a key role in molecular breeding for increasing crop yields. Although the rapid development of light detection ranging (LiDAR) provides new way to characterize three-dimensional (3D) structure, there is need develop robust algorithms extracting 3D phenotypic traits from LiDAR data assist gene identification selection. Accurate field environments remains challenging, owing difficulties segmentation organs individual plants...
Low-light image enhancement (LLIE) aims to recover illumination and improve the visibility of low-light images. Conventional LLIE methods often produce poor results because they neglect effect noise interference. Deep learning-based focus on learning a mapping function between images normal-light that outperforms conventional methods. However, most deep cannot yet fully exploit guidance auxiliary priors provided by in training dataset. In this paper, we propose brightness-aware network with...
Conventional spectral image demosaicing algorithms rely on pixels' spatial or correlations for reconstruction. Due to the missing data in multispectral filter array (MSFA), estimation of is inaccurate, leading poor reconstruction results, and these are time-consuming. Deep learning-based methods directly learn nonlinear mapping relationship between 2D mosaic images 3D images. However, focused only learning domain, but neglected valuable information frequency resulting limited quality. To...
Deep convolutional neural network (DCNN) based image codecs, consisting of encoder, quantizer and decoder, have achieved promising compression results. The major challenge in learning these DCNN models lies the joint optimization as well adaptivity to input images. In this paper, we proposed a architecture for compression, where decoder are jointly learned. Specifically, fully vector quantization (VQNet) has been quantize feature vectors representation, representative VQNet optimized with...
Feature selection is a process commonly used in machine learning. This paper examines two broad classes of feature methods: filter methods and wrapper to find their individual advantages disadvantages. selects different merits propose filter-Wrapper hybrid method (FWHM) optimize the efficiency selection. FWHM divided into phase, which orders these features according reasonable criterion at first, then select best based on final criterion. These experiments benchmark model engineering prove...
In recent years, researchers have become more interested in hyperspectral image fusion (HIF) as a potential alternative to expensive high-resolution imaging systems, which aims recover (HR-HSI) from two images obtained low-resolution (LR-HSI) and high-spatial-resolution multispectral (HR-MSI). It is generally assumed that degeneration both the spatial spectral domains known traditional model-based methods or there existed paired HR-LR training data deep learning-based methods. However, such...