- Remote Sensing in Agriculture
- Remote-Sensing Image Classification
- Remote Sensing and LiDAR Applications
- Land Use and Ecosystem Services
- Automated Road and Building Extraction
- Urban Heat Island Mitigation
- Advanced Image and Video Retrieval Techniques
- Forest ecology and management
- Urban Green Space and Health
- Remote Sensing and Land Use
- Species Distribution and Climate Change
- Plant Water Relations and Carbon Dynamics
- Tree-ring climate responses
- Medical Image Segmentation Techniques
- Advanced Image Fusion Techniques
- Fire effects on ecosystems
- 3D Surveying and Cultural Heritage
- 3D Shape Modeling and Analysis
- Atmospheric and Environmental Gas Dynamics
- Geochemistry and Geologic Mapping
- Forest Ecology and Biodiversity Studies
University of Toronto
2013-2020
Chinese Academy of Sciences
2018-2020
Aerospace Information Research Institute
2020
Sanya University
2018
Institute of Remote Sensing and Digital Earth
2018
Tokyo University of Science
2013
Xi'an Jiaotong University
2012
University of Tübingen
2004
Fine classification of vegetation types has always been the focus and difficulty in application field remote sensing. Unmanned Aerial Vehicle (UAV) sensors platforms have become important data sources various fields due to their high spatial resolution flexibility. Especially, UAV hyperspectral images can play a significant role fine types. However, it is not clear how ultrahigh react highly fragmented planting areas, variation will affect accuracy. Based on obtained from commercial imaging...
Image segmentation is an important process and a prerequisite for object-based image analysis, but segmenting into meaningful geo-objects challenging problem. Recently, some scholars have focused on hybrid methods that employ initial subsequent region merging since consider both boundary spatial information. However, the existing criteria (MC) only heterogeneity between adjacent segments to calculate cost of segments, thus limiting goodness-of-fit because homogeneity within should be treated...
Microsites related to microenvironmental conditions, including microclimate, seem be a key factor for the restoration of forests in subalpine area. Tree growth was studied Picea abies (L.) Karst. (Norway spruce) and Larix decidua Mill. (European larch) on 30 plots located at different microsites (i.e., elevations micro top o graphies combined) within zone (16801940 m) Schmirn Valley (Tyrol, Austria). The age trees 27 years larch 28 spruce. mean height biomass decreased significantly with...
The invasive emerald ash borer (EAB, Agrilus planipennis Fairmaire) infects and eventually kills endemic trees is currently spreading across the Great Lakes region of North America. need for early detection EAB infestation critical to managing spread this pest. Using WorldView-2 (WV2) imagery, goal study was establish a remote sensing-based method mapping undergoing various stages. Based on field data collected in Southeastern Ontario, Canada, an health score with interval scale ranging from...
The ability to quantify green vegetation across space and over time is useful for studying grassland health function improving our understanding of the impact land use climate change on grasslands. Directly measuring fraction cover labor-intensive thus only practical relatively smaller experimental sites. Remote sensing indices, as a commonly-used method large-area mapping, were found produce inconsistent accuracies when mapping in semi-arid grasslands, largely due mixed pixels including...
Delineating canopy gaps and quantifying gap characteristics (e.g., size, shape, dynamics) are essential for understanding regeneration dynamics understory species diversity in structurally complex forests. Both high spatial resolution optical light detection ranging (LiDAR) remote sensing data have been used to identify through object-based image analysis, but few studies quantified the pros cons of integrating LiDAR segmentation classification. In this study, we investigate whether...
Delineating individual tree crowns (ITCs) in high-spatial-resolution images can help to improve forest inventory and management. However, single-band watershed segmentation methods often fail delineate broadleaf species, particularly uneven-aged stands when a single-scale parameter is used fit segments reference of different sizes. In this study, we present multispectral multiscale fitting method for ITC delineation, the involves two steps: 1) produce subsequent fitting, which takes full...
A comprehensive forest resource inventory needs more detailed species information at individual tree level. Although conventional ground-based measurement fails to achieve this target in an efficient way, the emergence of high resolution remote sensing images has made it possible past decade. Individual crown delineation is one most critical steps for classification from images. However, still challenging delineate crowns deciduous forests because continuous canopy. In study, a multi-band...
AbstractOlgan larch is a traditional construction material used for the renovation of historical timber-frame buildings in China. However, acquiring necessary large-sized trees from old-growth forests has become challenge China because rare and inaccessible distribution these trees. In recent years, remote sensing imagery provided more effective alternative delineating tree crowns automatically with high accuracy. this study, an object-based method using Geoeye-1 developed. Tree crown...
Greenspace in urban areas is closely related to ecosystems, economy, culture, and society. Recently, rapid development expansion are always dominated by a series of human–environment interactions, which can lead various spatial patterns greenspace especially along the urban–rural gradient. Urban–rural mapping therefore great importance provide comprehensive insight for planners managers. In our study, we adopted both sub-pixel super-pixel strategies map Haidian District, Beijing, China....
Remote sensing image segmentation is the critical process in workflow of object-based analysis. Recently, region merging methods have attracted growing attention because they are able to utilize more features than spectral signals derived from initial segments. However, existing algorithms commonly use fixed parameters control merging, which limits possibility co-existence large and small To address this issue, we propose a self-adapted method, based on angle threshold, toward segmenting...
Recent advances in remote sensing technology provide sufficient spatial detail to achieve species-level classification over large vegetative ecosystems. In deciduous-dominated forests, however, as tree species diversity and forest structural increase, the frequency of spectral overlap between also increases our ability classify significantly decreases. This study proposes an operational workflow individual tree-based for a temperate, mixed deciduous using three-seasonal WorldView images,...
A novel variational level set model for multiple-building extraction from a single remote image is proposed in this paper. Multi-competing shapes are considered together with the model, curve evolution constrained by prior shape knowledge and label function that dynamically indicates region which should be compared. The building addressed through segmentation approach involves use of function, as well knowledge. In addition, permits translation, scaling, rotation shape. Experimental results...
Individual tree-based species maps are valuable for sustainable forest practices from both economic and ecological perspectives. Recent advances in high spatial resolution remote sensing provide the opportunity to map trees with greater accuracy. This study aims classify tree at individual level by using multi-seasonal WorldView-3 images. Our site is Haliburton Forest Wildlife Reserve, located Great Lakes-St. Lawrence region of Central Ontario, Canada. The images collected 2015 include late...
Point cloud semantic segmentation is a crucial task in 3D scene understanding. Existing methods mainly focus on employing large number of annotated labels for supervised segmentation. Nonetheless, manually labeling such point clouds the time-consuming. In order to reduce labels, we propose semi-supervised network, named SSPC-Net, where train network by inferring unlabeled points from few points. our method, first partition whole into superpoints and build superpoint graphs mine long-range...