Automated process for building material detection in remotely sensed imagery
A system and method for automatically (without human intervention) identifying a material in an image that comprises a building material for buildings in the image. Building side polygons which may be used to identify building sides in off-nadir imagery are generated. Off-nadir, multispectral images, building footprint data and elevation data for a geographic area are taken as input. Building heights for buildings in the geographic area are determined by clipping the elevation data using the building footprint data and then calculating building heights. A candidate set of polygons representing visible side faces of each building in the images is created from the known building heights, and based on the viewpoint direction, using vector analysis. After culling occluded polygons and polygons too small for analysis, the polygons are associated with a building footprint. Building materials for each building having visible polygons can then be identified.
1. An automated method for identifying a material in an image, comprising:
accessing multispectral spacecraft images for a geographic area, each image having a viewpoint direction;
accessing building footprint data and elevation data for the geographic area;
clipping the elevation data using the building footprint data to determine building heights;
determining polygons representing visible side faces of each building in the images based on the viewpoint direction using vector analysis, including culling ones of the one or more polygons which are occluded in the image to provide one or more visible polygons;
attaching ones of the one or more polygons representing visible side faces to a building footprint of a building in the geographic area; and
identifying materials in the polygons associated with each building.
2. The method of claim 1 wherein accessing building footprint data and structural elevation data comprises retrieving building footprint data from a shape file and retrieving elevation data in a raster from a digital surface model.
3. The method of claim 2 wherein clipping the elevation data comprises:
clipping the raster using the building footprint data to extract an interior area of each building to determine an interior elevation of each building;
clipping the raster using the building footprint data to extract an exterior border of the building to determine a ground elevation; and
calculating a building height of each building in the building footprint data as a maximum interior elevation of each building minus the ground elevation of each building.
4. The method of claim 1 wherein the clipping the elevation data is performed for each building before determining polygons representing visible side faces of each building.
5. The method of claim 1 wherein determining polygons representing visible side faces of each building comprises:
computing unit normal vectors for each side of each building; and
calculating visible side polygons of each building based on the viewpoint direction using vector analysis.
6. The method of claim 5 wherein the vector analysis comprises computing a dot product of a vector defining the viewpoint direction and each unit normal vector, and wherein a non-zero dot product indicates a visible face.
7. The method of claim 1 further including winnowing all visible faces by determining an order of polygons along the view direction and clipping those polygons more distant in the view direction based on a centroid computed for each building.
8. The method of claim 1 further including eliminating sliver polygons by removing those polygons smaller than a dimensional threshold and having a view angle above a threshold.
9. The method of claim 1 wherein the identifying comprises using a Spectral Angle Mapper and nominal spectrums for various types of building materials.
10. A computer implemented method of identifying a material in an image, comprising:
accessing multispectral spacecraft images, each image having a viewpoint direction, for a geographic area;
accessing a shape file having building footprint data for the geographic area;
accessing a digital surface model having a raster of structural elevation data for the geographic area;
clipping the structural elevation data using the building footprint data to determine building heights at building borders for each building in the geographic area;
determining polygons representing visible side faces of each building in the images based on the viewpoint direction using vector analysis, including culling ones of the one or more polygons which are occluded in the image to provide one or more visible polygons;
for each visible side faces in the geographic area, associating ones of the one or more polygons representing visible side faces to a building footprint of a building in the geographic area; and
identifying materials in the polygons associated with each building.
11. The method of claim 10 wherein clipping the structural elevation data comprises:
clipping the raster using the building footprint data to determine an interior elevation of each building;
clipping the raster using the building footprint data to determine a ground elevation at a building border; and
calculating a building height of each building in the building footprint data as the maximum of the interior elevation of said each building minus the ground elevation of said each building.
12. The method of claim 11 wherein determining polygons representing visible side faces of each building comprises:
computing unit normal vectors for each side of each building; and
calculating visible side polygons of each building based on the viewpoint direction using vector analysis.
13. The method of claim 12 wherein the vector analysis comprises computing a dot product of a vector defining the viewpoint direction and each unit normal vector such that each non-zero dot product represents a potentially visible polygon.
14. The method of claim 12 further including winnowing all visible faces by determining an order of polygons along the view direction and clipping those polygons more distant in the view direction based on a centroid computed for each building prior to said identifying.
15. The method of claim 14 further including eliminating sliver polygons by removing those polygons smaller than a dimensional threshold and having a view angle above a threshold prior to said identifying.
16. The method of claim 15 wherein the identifying comprises using a Spectral Angle Mapper and nominal spectrums for various types of building materials.
17. A non-transitory processor readable storage device having processor readable code embodied on the processor readable storage device, the processor readable code programming one or more processors to perform a method comprising:
accessing multispectral spacecraft images of a geographic area, each image having a viewpoint direction;
accessing building footprint data comprising one or more building footprints for buildings in the geographic area from a shape file;
accessing a raster of structural elevation data for the geographic area from a digital surface model;
for each building in the geographic area, clipping the structural elevation data using the building footprint data to determine building heights;
after clipping the structural elevation data for each building, determining polygons representing visible side faces of each building in the images based on the viewpoint direction using vector analysis, including culling ones of the one or more polygons which are occluded in the image to provide one or more visible polygons;
for each visible side faces in the geographic area, associating ones of the one or more visible polygons to a building footprint of a building in the geographic area; and
identifying materials in the polygons associated with each building.
18. The non-transitory processor readable storage device of claim 17 wherein clipping the structural elevation data comprises:
clipping the raster using the building footprint data to determine an interior elevation of each building;
clipping the raster using the building footprint data to determine a ground elevation at a building border; and
calculating a building height of each building in the building footprint data as the maximum of the interior elevation of said each building minus the ground elevation of said each building.
19. The non-transitory processor readable storage device of claim 18 wherein determining polygons representing visible side faces of each building comprises:
computing unit normal vectors for each side of each building; and
calculating visible side polygons of each building based on the viewpoint direction by computing a dot product of a vector defining the viewpoint direction and each unit normal vector.
20. The non-transitory processor readable storage device of claim 19 further including:
determining an order of polygons along the view direction and clipping those polygons more distant in the view direction based on a centroid computed for each building; and
eliminating sliver polygons by removing those polygons smaller than a dimensional threshold and having a view angle above a threshold prior to identifying.