National terrain mapping
The long-term objective of this research is to automatically extract and/or map terrain features for national mapping, and in so doing, set precedence for similar work in other subject matter realms.
The CEGIS 3DEP Initiative
The CEGIS 3DEP Initiative involves applications research projects, including pilots and test beds in areas such as the generation of derivative products from lidar that are National in scope, and creation of decision support systems with 3DEP and geospatial semantics.
The modeling, identification, and extraction mechanisms for terrain features such as mountains, hills, and valleys are in part dependent on an understanding of their creation, on their morphometric properties such as shape and size, and on naïve perception of the physical landscape.
Lidar data are being acquired as a part of the 3D Elevation Program (3DEP) and have sufficient resolution to capture the many and varied aspects of all types of terrain features. The ability to use these data as a source for extraction of geomorphologic and/or terrain features that can then be used to support spatial reasoning and natural language processing, and topographic science modeling and map generation, depends on a thorough understanding of both the features themselves and the everyday human conceptions of those features.
Landform Reference Ontology
The landform reference ontology (LFRO) formalizes land form categories, relationships and concepts. LFRO is intended to guide design and implementation of automated mapping and delineation of landforms from digital elevation models (DEMs) and other imagery.
The rationale and design implications of grounding LFRO in the upper level ontology DOLCE has led to many new categories and relationships. LFRO is also applied to explore nuances of common natural language landform terms and implications for their automated delineation. As a domain reference ontology, LFRO should serve as a neutral foundation and provide high-level abstraction of the (landform) domain. The choice of LFRO categories is based on qualitative 3D shape concepts because landforms are apprehended as unitary things due to their characteristic shape.
Topo Label Analysis using Machine Learning
New for FY2019, topo labels will be analyzed through machine learning techniques like Deep Convolutional Neural Networks (DCNN) in order to automatically extract map features and relocate feature annotation on USGS topographic maps of all scales.
CEGIS - Denver, Colorado
CEGIS - Rolla, Missouri
The long-term objective of this research is to automatically extract and/or map terrain features for national mapping, and in so doing, set precedence for similar work in other subject matter realms.
The CEGIS 3DEP Initiative
The CEGIS 3DEP Initiative involves applications research projects, including pilots and test beds in areas such as the generation of derivative products from lidar that are National in scope, and creation of decision support systems with 3DEP and geospatial semantics.
The modeling, identification, and extraction mechanisms for terrain features such as mountains, hills, and valleys are in part dependent on an understanding of their creation, on their morphometric properties such as shape and size, and on naïve perception of the physical landscape.
Lidar data are being acquired as a part of the 3D Elevation Program (3DEP) and have sufficient resolution to capture the many and varied aspects of all types of terrain features. The ability to use these data as a source for extraction of geomorphologic and/or terrain features that can then be used to support spatial reasoning and natural language processing, and topographic science modeling and map generation, depends on a thorough understanding of both the features themselves and the everyday human conceptions of those features.
Landform Reference Ontology
The landform reference ontology (LFRO) formalizes land form categories, relationships and concepts. LFRO is intended to guide design and implementation of automated mapping and delineation of landforms from digital elevation models (DEMs) and other imagery.
The rationale and design implications of grounding LFRO in the upper level ontology DOLCE has led to many new categories and relationships. LFRO is also applied to explore nuances of common natural language landform terms and implications for their automated delineation. As a domain reference ontology, LFRO should serve as a neutral foundation and provide high-level abstraction of the (landform) domain. The choice of LFRO categories is based on qualitative 3D shape concepts because landforms are apprehended as unitary things due to their characteristic shape.
Topo Label Analysis using Machine Learning
New for FY2019, topo labels will be analyzed through machine learning techniques like Deep Convolutional Neural Networks (DCNN) in order to automatically extract map features and relocate feature annotation on USGS topographic maps of all scales.
CEGIS - Denver, Colorado
CEGIS - Rolla, Missouri