Generating multi-dimensional building models with ground level images
A system and method is provided for generating textured building models from wirelessly captured ground-level imagery. Ground-level images of facades of building objects are collected for identification of geometry planes and generation of building façade geometry. The 3D building model is properly geo-positioned, scaled and textured.
1. A method of modeling planar surfaces, the method comprises:
receiving a plurality of ground-level images from a smartphone, the plurality of ground-level images corresponding to all sides of an entire building;
creating a point cloud with data points from the received plurality of ground-level images;
selecting a plurality of vertices defining edges of at least one single planar surface within the point cloud;
outlining the at least one single planar surface based on the selected plurality of vertices;
creating a simplified façade geometry for the at least one single planar surface based on the outlined single planar surface;
correlating data points within the point cloud to the outlined single planar surface;
averaging a planar fit for the outlined single planar surface based on the correlated data points; and
adjusting a position of the simplified façade geometry based on the average planar fit.
2. The method of claim 1 further comprises:
receiving an orthogonal image for the entire building;
identifying a plurality of orthogonal vertices of the orthogonal image;
correlating at least two of the plurality of orthogonal vertices with the adjusted simplified façade geometry; and
constructing a building model representing all surfaces of the entire building using at least the correlated plurality of orthogonal vertices.
3. The method of claim 1 further comprises applying a texture to the at least one planar surface.
4. The method of claim 3 , wherein the applying a texture comprises mapping at least one of the ground-level images to the at least one planar surface.
5. The method of claim 3 , wherein the applying a texture comprises using one or more of: camera information or ground-level image information.
6. A method of modeling planar surfaces, the method comprises:
receiving a plurality of ground-level images from a smartphone, the plurality of ground-level images corresponding to a plurality of sides of a building;
creating a three-dimensional (3D) point cloud with data points from the received plurality of ground-level images;
selecting a plurality of vertices defining edges of a plurality of planar surfaces within the 3D point cloud;
outlining the plurality of planar surfaces based on the selected plurality of vertices;
creating a simplified facade geometry for the plurality of sides of the building based on the outlined plurality of planar surfaces;
averaging a planar fit for the data points within the outlined plurality of planar surfaces; and
adjusting a position of at least one of the plurality of planar surfaces based on the average planar fit.
7. The method of claim 6 further comprises applying a texture to the plurality of planar surfaces.
8. The method of claim 7 , wherein the applying a texture comprises mapping at least one of the ground-level images to a corresponding planar surface of the plurality of planar surfaces.
9. The method of claim 7 , wherein the applying a texture comprises using one or more of: camera information or ground-level image information.
10. The method of claim 6 further comprises:
receiving an orthogonal image for an entire perimeter of the building;
identifying a plurality of orthogonal vertices of the orthogonal image;
correlating at least two of the plurality of orthogonal vertices with each of the plurality of planar surfaces; and
constructing a building model representing all surfaces of the building using at least the correlated plurality of orthogonal vertices.
11. A system for modeling planar surfaces, the system comprises:
a computer system for receiving, from a wireless capture device, a plurality of ground-level images of a building object captured by the wireless capture device circumnavigating a building object;
an image database for storing the received plurality of ground-level images of a building object;
an image processor, operative with the image database, to:
create a point cloud with data points from the received ground-level images;
select a plurality of vertices defining edges of at least one single planar surface within the point cloud;
outline the at least one single planar surface based on the selected plurality of vertices;
create a simplified facade geometry for the at least one single planar surface based on the outlined single planar surface;
correlate data points within the point cloud to the outlined single planar surface;
average a planar fit for the outlined single planar surface based on the correlated data points; and
adjust a position of the simplified façade geometry based on the average planar fit.
12. The system of claim 11 further comprises:
receiving an orthogonal image for the entire building;
identifying a plurality of orthogonal vertices of the orthogonal image;
correlating at least two of the plurality of orthogonal vertices with the adjusted simplified façade geometry; and
constructing a building model representing all surfaces of the entire building using at least the correlated plurality of orthogonal vertices.
13. The system of claim 11 further comprises applying a texture to the at least one planar surface.
14. The system of claim 13 , wherein the applying a texture comprises mapping at least one of the ground-level images to the at least one planar surface.
15. The system of claim 13 , wherein the applying a texture comprises using one or more of: camera information or ground level image information.
16. The system of claim 11 further comprises a three dimensional (3D) proximity buffer applied to selected data points in the point cloud to determine the at least one planar surface and identify data points in the point cloud that are in close proximity to a region around each of the selected data points.
17. The system of claim 16 , wherein a closest actual data point to the region is used as the data point to determine the at least one planar surface.