详情

Fusing multi-source UAS-derived data to improve project planning and the NCDOT Wetlands Prediction Model

项目名称：	Fusing multi-source UAS-derived data to improve project planning and the NCDOT Wetlands Prediction Model
摘要：	Since the UNCW Socio-Environmental Analysis Lab (SEAL) directed by Dr. Pricope has acquired the professional mapping drone SenseFly eBee Plus RTK at the end of 2016, we have been developing various unmanned aerial systems (UAS) applications and data collection protocols, processing workflows and UAS-based remotely acquired data integration using multiple types of sensors, including very high resolution visual/RGB, multispectral and thermal. Since then, we have additionally acquired off-the-shelf multi-rotor quadcopters (various DJI models) that have been used in combination with multispectral sensors (specifically, the SEAL lab owns the high-end multispectral camera Parrot Sequoia with green, red, red edge and near-infrared sensors) for challenging missions that require vertical take-off and landings (closed canopies, edge of water bodies, very heterogeneous vegetation cover). More recently, under the umbrella of a newly-created UNCW Geospatial Service Group, in partnership with UNCW Center for Innovation and Entrepreneurship, we have been conducting extensive UAS-based research and development centered on: establishing field data collection, sampling and calibration and validation protocols for effective UAS data collection and processing using the three different types of drone sensors (plus a proposed drone-borne LiDAR sensor if the request should be approved); implementing processing workflows to create validated, ortho-photogrammetrically and planimetrically correct UAS20 derived products (orthomosaics, reflectance maps, indices, 3D digital surface and digital terrain models, 3D models and visualizations, derived metrics such as canopy texture, heights, sizes); and applying and developing geospatial analysis and UAS to satellite imagery fusion techniques. In this proposal, we highlight our capabilities and present methods in which UNCW’s expertise in designing and collecting UAS data, remote sensing fusion and image processing techniques to derive vegetation characteristics, and geospatial analysis techniques to classify and derive wetland and other sensitive habitats will be seamlessly utilized to provide additional support in the development of habitat mapping for NC DOT. Specifically, we propose to leverage our extensive field methods and UAS data processing experience to demonstrate the utility of incorporating 1) hyper-resolution RGB, 2) LiDAR (with derived metrics), 3) multispectral, and 4) thermal imagery. Imagery/data will be collected with either a fixed wing or a quadcopter drone along with ancillary data such as RTK GPS ground control points (GCPs) and spectral curves derived from a field Spectroradiometer. Incorporating the field data and UAS data enables us to create highly-resolved models for sampling sites selected throughout the three ecoregions in North Carolina (Figure 1). At selected sites in the three ecoregions (chosen in collaboration with NC DOT), using our established data collection protocols that include GCPs and vegetation species spectral curve recordings, our team will collect four sets of imagery (RGB, multispectral, thermal and LiDAR). Data will be processed using the high-end Pix4D software into various derived indices and metrics and then classified into land cover objects based on spectral, textural, elevation and other measures using object-based classifications in eCognition. After UAS-based classifications have been completed, accuracy assessments will be conducted utilizing field collected GPS data points and randomly generated points in the larger NC DOT project sites. Next, we will use the objects derived from the drone imagery along with the reference samples collected in the field (including species spectra) to train satellite imagery to classify vegetation species over larger, regional scales using several advanced classification algorithms, ancillary geospatial datasets and R statistical modeling software. This will establish spatial extents for various vegetation (wetlands and other habitats) species in the three representative ecoregions and allow us to derive very high resolution land cover maps to be used as input into the Wetland Predictive Modeling software currently being used at NC DOT, thus leading to significant time and cost savings for NC DOT projects.
状态：	Active
资金：	300,411
资助组织：	North Carolina Department of Transportation
项目负责人：	Kirby, John
执行机构：	University of North Carolina at Wilmington
开始时间：	20190801
预计完成日期：	20210731
主题领域：	Environment