Progress 10/01/11 to 09/30/16
Outputs
Target Audience:Land use planners, managers, and decision makers in Connecticut. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?The project supported a PhD level graduate assistant who was trained on data integration and processing techniques for improving land use information. How have the results been disseminated to communities of interest?Project results have been disseminated through peer reviewed journals, conference presentations and a webinar. What do you plan to do during the next reporting period to accomplish the goals?
Nothing Reported
Impacts
What was accomplished under these goals?
Agriculture land-use information is important for supporting agricultural policies, verifying claims by farmers who apply for subsidies, and ensuring compliance with government sponsored farm programs. High spatial resolution National Agriculture Inventory Program (NAIP imagery (i.e. 1- meter) used in conjunction with high temporal resolution (16 days) Landsat Operational Land Imager (OLI) imagery has the potential to create 1-meter resolution crop land maps in Connecticut. A semi-automated, object-based approach was developed to classify agricultural land use by leveraging the high spatial resolution of NAIP and the high temporal resolution of Landsat. The research focused on classifying six common agricultural crop types in Connecticut including corn, alfalfa, shaded tobacco, nursery stock, pumpkin, and other (e.g. turf). A Mean Shift segmentation algorithm was used to extract low-height objects from a LiDAR-derived Digital Height Model (DHM). For the low-height objects, vegetated objects were identified using the Normalized Difference Vegetation Index (NDVI) derived from NAIP imagery. A temporal series of cloud-free Landsat OLI images were used to assess the phenology of the vegetated objects based on the NDVI profiles over the course of the year. The phenological properties along with other characteristics of the objects, such as spectral, spatial and contextual information, were used to develop a hierarchical decision tree to classify agricultural land use. The decision tree was applied to three sample towns in eastern Connecticut which have substantial agricultural land areas with target crop types. Qualitative accuracy assessments showed the method has good promise and a rigorous quantitative assessment will be used to further evaluate the method's performance. In addition to agriculture land-use mapping, progress has been made in shadow detection and classification of features of shadows in high resolution aerial imagery. In previous research, we developed a mathematical model to predict shadow location in high spatial resolution aerial imagery based on solar geometry and building heights using LiDAR-derived height model. However, the shadow location predicted from the model problem did not match with the actual shadow location in the aerial imagery due to the parallax of the buildings. To address this problem, the potential shadow pixels in the aerial imagery were classified using unsupervised method based on spectral properties. An exhaustive computer algorithm was developed to align the predicted shadow location to actual shadow location in the imagery deriving final shadow location. Once shadows had been successfully identified, radiometric enhancement along with an object-based classification approach was used to classify land cover in the shadow areas. Linear correlation correction (LCC) and histogram matching (HM) methods were performed on the final shadow areas to test their ability to improve land cover classification. The aerial imagery within the shadows are extracted and segmented into objects using the Mean Shift segmentation algorithm. A decision tree was developed to classify land cover features in the shadow areas based on the shadow objects' spectral, spatial and contextual characteristics. Results summary for the past 5 years The overall research goals of this project aimed to address the challenge of deriving high resolution, thematically-detailed land cover and land use information for a large geographic extent (e.g. statewide) using multi-source geospatial data including LiDAR, satellite imagery, aerial imagery as well as ancillary dataset such as parcel boundaries. The accomplishments from this research are the: Development of a predictive model for shadow detection in high spatial resolution aerial imagery based on the solar geometry and building heights. Classification of land-use types for building structures, including single-family residential, multi-family residential, non-residential, using high spatial resolution land-cover data with parcel boundaries. Classification of cropland types and other grassy vegetation (e.g. turf) using high spatial and high temporal resolution remote sensing data.
Publications
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2016
Citation:
Lei, Q. 2016. Mapping agricultural land use by integrating high resolution remote sensing imagery with multi-temporal Landsat data. ASPRS 2016. Fort Worth, TX. April 2016.
- Type:
Journal Articles
Status:
Other
Year Published:
2017
Citation:
In Preparation. Lei, Q. 2017. Land use classification for building structures using multi-source spatial data.
|
Progress 10/01/14 to 09/30/15
Outputs
Target Audience:Land use planners, managers, and decision makers in Connecticut. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?
Nothing Reported
How have the results been disseminated to communities of interest?Project results have been disseminated through peer reviewed journals, conference presentations and a webinar. What do you plan to do during the next reporting period to accomplish the goals?Over the next period, we will continue to improve our shadow classification algorithm by addressing the issue of parallax in building rooftops. We plan to use flightline data, for the image acquisition mission, along with digital surface models to predict where rooftops can be expected to have been shifted in an image. If parallax can be successfully identified through an automated process, then rooftops can be removed from the shadow classification thereby allowing the shadows to be spectrally enhanced and improving classification of shaded features. In the near future, we will complete development of the road model and update the high resolution land cover map as well as use the model to differentiate land use for impervious features. Additionally, we will develop new methods to distinguish among natural grass, agriculture, and turf lands using time series of Landsat data. Although these land uses have very similar spectral properties during summer time, we expect they will have unique spectral signatures when accounting for different times of year. If successful, the grass land use classifications would be used to update our high resolution land cover/use map. In the past, LiDAR data covering the state of Connecticut have been collected in a "piece meal" basis, consisting of 11 different missions spanning more than 10 years in order to cover the state fully. Since our study areas are from different areas of the state, we have had to deal with the different spatial and temporal properties of LiDAR data and their derivatives used in our research. In spring 2016, a mission is scheduled to produce temporally- and spatially-consistent high resolution LiDAR, as well as 3-inch resolution four-band orthoimagery. Once we acquire these data, we will adapt procedures developed in the project to the new LiDAR and orthoimagery to facilitate transfer of the protocols to any location in the state, thereby establishing the foundation of producing statewide high resolution land use data in a future project.
Impacts
What was accomplished under these goals?
Shadows are problematic in high resolution imagery and are a significant source of classification errors particularly during leaf-on conditions and in areas with tall buildings. Classification of shadows in an image would allow shaded features to be spectrally enhanced and improve classification accuracy. Over the past year, we have made progress in classifying shadows using a sunlight geometry model along with a digital surface model derived from airborne Light Detection and Ranging (LiDAR) data. We developed an algorithm that automates the construction of the sunlight geometry model based on the image acquisition time and the heights of features in the digital surface model. One problem encountered in the classification of shadows is due to the parallax in the aerial imagery - shadows predicted from the sunlight geometry model are often partially obscured by building rooftops. Further work is needed to develop methods to identify where parallax is confounding the classification of shadows. In addition to shadow detection, we have made progress in improving the road classification in the 8-class high resolution land cover map that we developed in previous research. The classification algorithm used to create this land cover map was unable to classify roads overshadowed by coniferous trees or dense deciduous trees. Recently, we received accurate road centerline data for the entire state as well as planimetric data for 8 towns in southwestern Connecticut. We used the road centerlines and planimetric data to calculate the widths of roads and found that speed limits are good predictors of road widths for most classes of road. The statewide road centerline data includes speed limits for all roads and can be used to model road widths throughout the state. In future work, we will develop this road model and use it to update and correct our high resolution land cover map. The road model will further be used to distinguish roads from parking lots in the land cover map.
Publications
- Type:
Journal Articles
Status:
Published
Year Published:
2015
Citation:
Parent, J., J. Volin, and D.L. Civco. 2015. A fully-automated approach to land cover mapping with airborne LiDAR and high resolution multispectral imagery in a forested suburban landscape. ISPRS Journal of Photogrammetry and Remote Sensing 104(2015):18-29. DOI: 10.1016/j.isprsjprs.2015.02.012.
- Type:
Journal Articles
Status:
Published
Year Published:
2014
Citation:
Bentley, G.C., D.M. Hanink, R.G. Cromley, C. Zhang and D.L. Civco. 2014. Analyzing Open Space Distributions in the Context of the Environmental Kuznets Curve: An Example from the Northeastern United States. The Northeastern Geographer 6(2014):1-26.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2014
Citation:
Civco, D.L., S. Angel, J.D. Hurd., J. Parent and A. Chabaeva. 2014. Atlas of Urban Expansion 2015 Edition. Pecora 19: "Sustaining Land Imaging: UAS to Satellites" in conjunction with The Joint Symposium of ISPRS Technical Commission I & IAG Commission 4. Denver. CO. November 2014.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2014
Citation:
Parent, J. and Q. Lei. 2014. A Fully Automated Approach to Classifying Urban Land Use and Cover from LiDAR, Multi-spectral Imagery, and Ancillary Data. Pecora 19: "Sustaining Land Imaging: UAS to Satellites" in conjunction with The Joint Symposium of ISPRS Technical Commission I & IAG Commission 4. Denver. CO. November 2014.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2014
Citation:
Lei, Q. 2015. A Fully Automated Approach to Classifying Urban Land Use and Cover from LiDAR, Multi-spectral Imagery, and Ancillary Data. UConn College of Agriculture, Health, and Natural Resources Graduate Student Forum, Storrs, CT. March 2015.
- Type:
Other
Status:
Published
Year Published:
2015
Citation:
Angel, S., A.M. Blei, D.L. Civco, N.G, Sanchez, P. Lamson-Hall, M.Madrid, J. Parent, and K.Thom1. 2015. Monitoring the Quantity and Quality of Global Urban Expansion. Marron Institute Working Paper #24. (http://marroninstitute.nyu.edu/content/working-papers/monitoring-the-quantity-and-quality-of-global-urban-expansion). 25 p.
|
Progress 10/01/13 to 09/30/14
Outputs
Target Audience: Land use planners, managers, and decision makers in Connecticut. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?
Nothing Reported
How have the results been disseminated to communities of interest? Conference presentations (NEARC 2014 and ASPRS Pecora 2014). One peer-reviewed journal manuscript in review. One PhD dissertation in preparation. What do you plan to do during the next reporting period to accomplish the goals? More rule sets will be developed to extract other detailed land use features (e.g. agriculture versus turf, urban- versus non-urban forests). Accuracy assessment for the developed urban land use classification rule sets will be performed on three Connecticut towns in each of three development classes - urban (Bridgeport, Manchester, Greenwich), suburban (Mansfield, Redding, Plainfield), and rural (Woodstock, Hampton, Franklin). Shadows cause problems in land cover classification from high resolution imagery. The land cover features in shadow areas are often misclassified or left unclassified, which can result in significant loss of land cover information. With the wide availability of high resolution imagery, there is increasing interest in extracting and correcting shadows caused by features on the ground. Previous research has relied on hyperspectral remote sensing data which is not commonly accessible, particularly for large extents. We propose to develop an automated approach that combines the sunlight geometry model and ray tracing technique to detect shadow areas, integrating high resolution multispectral remote sensing imagery and airborne LiDAR data. Accuracy assessment will be performed to evaluate the effectiveness of the proposed approach.
Impacts
What was accomplished under these goals?
The increasing availability of high resolution remote sensing data provides more opportunities to derive spatially and thematically enhanced land cover and land use information for large geographical extents. In this project, eight land cover types (buildings, low impervious cover, water, riparian wetlands, deciduous forests, conifer forests, low vegetation, and medium vegetation) were classified, integrating high resolution LiDAR data and multispectral orthophotographs. The classification algorithm used pixel- and object-based rulesets, based on features' structural and spectral properties, and was implemented in a Python script. The accuracy assessment for all eight land cover types was performed and indicated that the proposed rulesets were robust across the range of landscapes in eastern Connecticut. Three urban land use types (non-residential, single family residential and multi-family residential) were derived from the urban land cover classification (i.e., buildings) and Connecticut parcel data. The classification algorithm, based on building spatial (shape and size) and contextual information (parcel land cover composition), was implemented in a Python script. Based on visual assessments, the preliminary rule set showed promise for classifying land use of buildings. The overall accuracy of the urban land covers (buildings, low impervious cover) was 95%.
Publications
- Type:
Journal Articles
Status:
Published
Year Published:
2014
Citation:
Zhang. C., W. Lei, and D. Civco. 2014. Application of Geographically Weighted Regression to gap-fill of SLC-off Landsat ETM+ satellite imagery. International Journal of Remote Sensing. 35(22):7650-7672. DOI:10.1080/01431161.2014.975377.
- Type:
Journal Articles
Status:
Published
Year Published:
2014
Citation:
Zhang D, C. Zhang, W. Li, R.G. Cromley, D. Hanink, D.L. Civco, and D. Travis. 2014. Restoration of the missing pixel information caused by contrails in multispectral remotely sensed imagery. Journal of Applied Remote Sens. 0001;8(1):083698. doi:10.1117/1.JRS.8.083698.
- Type:
Journal Articles
Status:
Published
Year Published:
2014
Citation:
Witharana C. and D.L. Civco. 2014. Optimizing multi-resolution segmentation scale using empirical methods: Exploring the sensitivity of the supervised discrepancy measure Euclidean distance 2 (ED2). ISPRS Journal of Photogrammetry and Remote Sensing 87(1):108-121. doi:10.1016/j.isprsjprs.2013.11.006.
- Type:
Journal Articles
Status:
Published
Year Published:
2014
Citation:
Witharana C., D.L. Civco, and T.H Meyer. 2014. Evaluation of data fusion and image segmentation in earth observation based rapid mapping workflows. ISPRS Journal of Photogrammetry and Remote Sensing 87(1):1-18. doi:10.1016/j.isprsjprs.2013.10.005.
- Type:
Conference Papers and Presentations
Status:
Other
Year Published:
2014
Citation:
Parent, J. and Q. Lei. 2014. A Fully Automated Approach to Classifying Urban Land Use and Cover from LiDAR, Multi-spectral Imagery, and Ancillary Data. Presented at the Northeast Arc User Group (NEARC) Conference, Mystic, CT.
- Type:
Journal Articles
Status:
Under Review
Year Published:
2015
Citation:
Parent, J., J. C. Volin, and D.L. Civco. 2015. A fully-automated approach to land cover mapping with airborne LiDAR and high resolution multispectral imagery in a forested suburban landscape. ). ISPRS Journal of Photogrammetry and Remote Sensing. (in review)
|
Progress 01/01/13 to 09/30/13
Outputs
Target Audience: Planners and decision makers at the municipal and state level widely use land use and land cover data. These data are used extensively, too, in research and outreach applications. To date, only moderate spatial and thematic resolution statewide land cover information, based on Landsat satellite remote sensing data, has been available, As noted by the LandCoverandLand Use Subcommittee of the Data Inventory and Assessment Working Group of the Connecticut Geospatial Information Systems Council (CGISC), users at all levels of government have requested land use information at a finer spatial scale and with more detailed classes, especially for urban categories. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided? Preliminary data have been used in classroom demonstrations for several geomatics courses in the Department of Natural Resources and the Environment at the University of Connecticut. How have the results been disseminated to communities of interest?
Nothing Reported
What do you plan to do during the next reporting period to accomplish the goals? More rule sets will be developed to extract other detailed land use features. Accuracy assessment for the developed urban land use classification rule sets will be performed on three towns in each of three development classes – urban (Bridgeport, Manchester, Greenwich), suburban (Mansfield, Redding, Plainfield), and rural (Woodstock, Hampton, Franklin). In addition to our land cover and land use classification rule sets, we propose to develop a transferable and adaptable, object-based classification scheme using fuzzy rule sets for areawide land cover and land use mapping through the synergistic use of remote sensing and ancillary data. The procedures will be implemented in an eCognition workflow, and the corresponding eCognition Architect Solution will be generated, enabling a high degree of interoperability. An assessment approach will be performed to quantify the robustness of the classification scheme by reapplying the scheme to different but comparable areas and data sources. Ultimately, a Connecticut statewide, hierarchical, spatially and thematically enhanced land cover and land use map will be produced.
Impacts
What was accomplished under these goals?
Mapping urban land use features is essential for many applications. While land cover is typically classified by the physical characteristics from remote sensing data alone, land use requires more contextual information related to social and economic elements. Ancillary data, such as census and parcel information, have been used successfully in land use classification of urban areas. Census Bureau population data have been used to assist in distinguishing different classes of residential and non-residential land uses. Parcel spatial data have been used to provide cadastral information which can assist in the determination of land use. An automated approach for classifying urban land cover features (i.e. buildings) into detailed land use classes, integrating high-resolution remote sensing data, as well as census block and parcel data, was developed. A knowledge-based decision-tree approach was implemented in a Python script to derive four detailed land use classes for urban structures - single-family, multi-family, institutional and other non-residential). We aimed at generating a hierarchical, spatially-enhanced and thematically more detailed land cover and land use information at extended geographic extents (e.g., statewide) using multisource spatial data.
Publications
- Type:
Journal Articles
Status:
Published
Year Published:
2013
Citation:
Lin J., R. Cromley, D. L. Civco, D. Hanink, and C. Zhang. 2013. Evaluating the Use of Publicly Available Remotely Sensed Land Cover Data for Areal Interpolation. GIScience & Remote Sensing. Published online: 14 Jun 2013. DOI:10.1080/15481603.2013.795304.
- Type:
Book Chapters
Status:
Awaiting Publication
Year Published:
2014
Citation:
Zimmerman, C. and D. L. Civco. Impervious Surface Area and Their Effects. Encyclopedia of Natural Resources. Taylor & Francis. (Accepted September 2013, to be published June 2014)
- Type:
Journal Articles
Status:
Published
Year Published:
2013
Citation:
Allen, J.M., T. J. Leininger, J. D. Hurd Jr., D. L Civco, A. E. Gelfand, and J. A. Silander Jr. 2013. Socioeconomics drive woody invasive plants in New England through forest fragmentation. Landscape Ecology 28(9):1671-1686.
|
Progress 01/01/12 to 12/31/12
Outputs
OUTPUTS: A model has been developed to derive bare earth digital elevation model (DEM) data and feature heights above local terrain from airborne LiDAR data. This new approach to deriving a bare earth DEM in areas with voids (gaps) in the LiDAR point data set uses rule-based filtering algorithms as well as ancillary sources of elevation data (National Elevation Data, or NED) to "fill in" those gaps. In related research, an approach for quantifying deciduous forest canopy closure using leaf-off LiDAR data has been developed. These LiDAR-derived data are being used as part of the multisource geospatial information approach to land use and land cover classification and will be made available for publication use via Connecticut Environmental Conditions Online (http://www.cteco.uconn.edu/index.htm). These models will be brought to the attention of the science community by way of technical papers, as well as making them available to the GIS user community by way of the Center for Land use Education and Research (http://clear.uconn.edu). PARTICIPANTS: Daniel Civco, PI and James Hurd Co-PI. The project supports part of a PhD-level graduate assistant, whose principal research role is to explore innovative data integration and processing techniques for improving land use information. Not formally a member of the project team, but involved in related research, another PhD student has been instrumental in developing our LiDAR data processing models. Other participants include the end-users of the data, who have been surveyed about their thematic and spatial resolution needs in their applications of land use and land cover information (http://www.surveymonkey.com/s/T3JHQBW). TARGET AUDIENCES: The research itself is of value to the remote sensing community involved with area-wide land use and land cover mapping. The final statewide land use maps will be use to land planners, decision-makers, and researchers who require quality land use and land cover information. The Center for Land use Education and Research, its mission being to provide information, education and assistance to land use decision makers, in support of balancing growth and natural resource protection, will be the principal delivery portal for new land use data, as it is now for its Connecticut's Changing Landscape project (http://clear.uconn.edu/projects/landscape/index.htm). PROJECT MODIFICATIONS: The study areas for developing and testing advanced land use and land cover classification algorithms have been modified to take advantage of available remote sensing and other geospatial data. Three towns in each of three development classes - urban (Bridgeport, Manchester, Greenwich), suburban (Mansfield, Redding, Plainfield) , and rural (Woodstock, Hampton, Franklin) - have been selected. The use of RapidEye five-meter satellite remote sensing data in a multiresolution segmentation and classification approach has been de-emphasized because of the statewide availability of recent, high resolution, springtime (leaf-off) and late summertime (leaf-on), four-band aerial digital imagery.
Impacts
The incorporation of feature height data, coupled with other remote sensing data, has proven to be extremely important to the extraction of certain types of land use types, most notably residential houses and commercial buildings. Furthermore, multitemporal remote sensing data has facilitated the classification of land cover types that possess seasonal variations in their reflectance properties. Another unique aspect of our approach to extracting feature height data from LiDAR is to identify and remove automatically spurious values, introduced by flocks of birds for example. In addition to our land use and land cover classification rulesets, we aim at developing a new spatial hierarchical data structure for land use and land cover information, one in which the appropriate level of thematic and spatial detail as a function of geographic scale and scope is portrayed.
Publications
- Angel, S., J. Parent and D.L. Civco. 2012. The fragmentation of urban landscapes: global evidence of a key attribute of the spatial structure of cities, 1990 to 2000. Environment and Urbanization 24(1):249-283.
- Civco, D.L. and C. Witharana. 2012.Assessing the spatial fidelity of resolution-enhanced imagery using Fourier analysis: a proof-of-concept study, Proc. SPIE 8538, Earth Resources and Environmental Remote Sensing/GIS Applications III, 853805 (October 25, 2012); doi:10.1117/12.974703; http://dx.doi.org/10.1117/12.974703.
- Witharana, C. and D.L. Civco 2012. Evaluating remote sensing image fusion algorithms for use in humanitarian crisis management, Proc. SPIE 8538, Earth Resources and Environmental Remote Sensing/GIS Applications III, 853807 (October 25, 2012); doi:10.1117/12.973745; http://dx.doi.org/10.1117/12.973745.
- Witharana, C., D.L. Civco, and T.H. Meyer. 2013. Evaluation of pansharpening algorithms in support of earth observation based rapid-mapping workflows. Applied Geography 37(Feb):63-87. [Available on-line December 2012]
|
|