3D view model generation of an object utilizing geometrically diverse image clusters

1. A computer vision method, executed by one or more processors, for generating a single three-dimensional (3D) model view of a geographic scene, the method comprising:

receiving image data for the scene from a plurality of sensors located at different angles with respect to the geographic scene;

dividing the image data into a plurality of image spatial regions;

correlating the image data in each image spatial region to obtain a score for each image data in each image spatial region;

grouping the image data in each image spatial region into two or more image clusters, based on the scores for each image;

performing a multi-ray intersection within each image cluster to obtain a 3D reference point for each region;

for each region, combining the one or more clusters, based on the 3D reference point for the region; and

registering the combined clusters for each region to obtain a single 3D model view of the scene, wherein registering the combined clusters comprises:

rasterizing patches in two or more point clouds derived from the two or more image clusters;

determining 3D tie-points between the two or more point clouds via cross correlation processing;

performing a 3D bundle adjustment to determine misregistration error model parameters between the two or more point clouds;

applying the misregistration error model parameters to original point clouds, resulting in registered point clouds; and

merging the registered point clouds, resulting in a single, dense point cloud comprising of point clouds from each cluster.

2. The method of claim 1 , wherein the plurality of sensors are located on one or more platforms.

3. The method of claim 1 , wherein the plurality of image spatial regions are obtained by grouping the image data into coordinal and inter-coordinal map locations.

4. The method of claim 1 , wherein the image data is divided into the plurality of image spatial regions by selecting a reference image, finding correlation sets with higher strength that have diversity of convergence angles and ensuring image spatial regions have the largest convergence angles from all other groups.

5. The method of claim 1 , wherein correlating the image data further comprises geometrically analyzing each image spatial region and photogrammetrically analyzing each image spatial region to obtain a score for each image data in each image spatial region.

6. The method of claim 1 , wherein the score indicates a predicted quality of a point cloud generated from a corresponding image spatial region.

7. The method of claim 1 , wherein the multi-ray intersection utilizes a triangulation method.

8. The method of claim 1 , wherein the multi-ray intersection generates dense tie-points as a set of 1-to-N tie-points that resolve to an XYZ location in a 3D space to generate a point cloud from a reference image's perspective with a resolution half as dense as an original reference image.

9. The method of claim 1 , wherein combining the two or more clusters comprises merging multiple point clouds from each cluster perspective into a single point cloud representing the full scene.

10. The method of claim 1 , wherein correlating the image data utilizes a normalized-cross correlation score.

11. A computer vision system for generating a single three-dimensional (3D) model view of a geographic scene comprising:

one or more processors;

storage medium coupled to the one or more processors; and

I/O circuits, wherein the one or more processors:

receive image data for the scene from a plurality of sensors located at different angles with respect to the geographic scene;

divide the image data into a plurality of image spatial regions;

correlate the image data in each image spatial region to obtain a score for each image data in each image spatial region;

group the image data in each image spatial region into two or more image clusters, based on the scores for each image;

perform a multi-ray intersection within each image cluster to obtain a 3D reference point for each region;

for each region, combine the one or more clusters, based on the 3D reference point for the region; and

register the combined clusters for each region to obtain a single 3D model view of the scene, wherein registering the combined clusters comprises:

rasterizing patches in two or more point clouds derived from the two or more image clusters;

determining 3D tie-points between the two or more point clouds via cross correlation processing;

performing a 3D bundle adjustment to determine misregistration error model parameters between the two or more point clouds;

applying the misregistration error model parameters to original point clouds, resulting in registered point clouds; and

merging the registered point clouds, resulting in a single, dense point cloud comprising of point clouds from each cluster.

12. The system of claim 11 , wherein the plurality of sensors are located on one or more platforms.

13. The system of claim 11 , wherein the plurality of image spatial regions are obtained by grouping the image data into coordinal and inter-coordinal map locations.

14. The system of claim 11 , wherein the processor divides the image data into the plurality of image spatial regions by selecting a reference image, finding correlation sets with higher strength that have diversity of convergence angles and ensuring image spatial regions have the largest convergence angles from all other groups.

15. The system of claim 11 , wherein the processor correlates the image data by geometrically analyzing each image spatial region and photogrammetrically analyzing each image spatial region to obtain a score for each image data in each image spatial region.

16. The system of claim 11 , wherein the score indicates a predicted quality of a point cloud generated from a corresponding image spatial region.

17. The system of claim 11 , wherein the processor performs the multi-ray intersection to generate dense tie-points as a set of 1-to-N tie-points that resolve to an XYZ location in a 3D space to generate a point cloud from a reference image's perspective with a resolution half as dense as an original reference image.

18. The system of claim 11 , wherein the processor combines the two or more clusters by merging multiple point clouds from each cluster perspective into a single point cloud representing the full scene.