Computer vision database platform for a three-dimensional mapping system
A system is provided including a database that ingests data from disparate image sources, with a variety of image metadata types and qualities, and manages images geospatially through the creation and continued refinement of camera solutions for each data object included. These camera solutions are calculated and refined by the database as additional data enters the system that could affect the solutions, through a combination of the application of image metadata towards image processing methods and the use of optical-only computer vision techniques. The database continually generates data quality metrics and relevant imagery and geometry analytics, which drive future collection tasking, system analytics, and human quality control requirements.
1. A method of updating building model textures comprising:
receiving one or more collected images, the one or more collected images including one or more facades of a building;
processing regions of the one or more collected images to identify architectural features of a selected façade within the regions as metadata associated with a respective one of the one or more collected images;
comparing at least one of the one or more collected images to a plurality of previously stored geospatial images, the comparing including an analysis of the identified architectural features metadata;
ranking the comparing of the respective one of the one or more collected images and the plurality of previously stored geospatial images, the ranking producing a ranked list of similar images to the respective one of the one or more collected images based at least on the analysis of the metadata;
creating a pixel-to-pixel correlation between the respective one of the one or more collected images and a highest ranked of the plurality of previously stored geospatial images;
storing collected pixel-to-pixel correlated images in a database with metadata data derived during both the processing as well as with metadata associated with the highest ranked of the plurality of previously stored geospatial images; and
when the pixel-to-pixel correlation indicates differences between the respective one of the one or more collected images and the highest ranked of the plurality of previously stored geospatial images, updating the building model textures by retexturing at least one of the facades of a building model with the pixel-to-pixel correlated one of the at least one of the one or more collected images.
2. The method according to claim 1 , wherein the database is queried by any of: computer vision algorithms or image processing algorithms.
3. The method according to claim 1 , wherein the database accepts any of: orthographic, oblique aerial, satellite, terrestrial vehicular or crowdsourced terrestrial imagery.
4. The method according to claim 1 , wherein the one or more collected images are processed to identify, analyze, or remove obstructions before comparing.
5. The method according to claim 1 , wherein a match probability metric of the at least one of the one or more collected images and the plurality of previously stored geospatial images are ranked according to a weighting criteria.
6. The method according to claim 1 , wherein the highest ranked of the ranked list of similar images is validated.
7. The method as according to claim 5 , wherein the match probability metric includes selection criteria.
8. The method as according to claim 7 , wherein the selection criteria are used to adjust future weighting criteria for ranking.
9. A system for updating of building model textures comprising:
an image acquisition module receiving a collected image from a capture device;
a database for processing and storing geospatial images;
a query analysis module receiving the collected image from the image acquisition module and processing to remove obstructions and analyzing the collected image to identify building architectural features within a selected façade of the collected image;
a comparison metrics module for comparing the analyzed collected image against a plurality of the stored geospatial images with similar building architectural features to the identified building architectural features;
a ranking module for ranking the collected image against the plurality of the stored geospatial images, based on the comparing;
a pixel-to-pixel correlation module that creates, for a highest ranked one of the plurality of the stored geospatial images, a pixel-to-pixel correlation between the collected image and a corresponding one of the plurality of the stored geospatial images; and
a texturing module to retexture one or more facades of a corresponding building model including the pixel-to-pixel correlation with the collected image.
10. The system according to claim 9 , wherein the database is queried by any of: computer vision algorithms or image processing algorithms.
11. The system according to claim 9 , wherein the database accepts orthographic, oblique aerial, satellite, terrestrial vehicular or crowdsourced terrestrial imagery.
12. The system according to claim 9 , wherein the database accepts imagery from a camera system.
13. The system according to claim 9 , wherein a match probability metric of the collected image and the plurality of the stored geospatial images includes ranking according to a weighting criteria.
14. The system according to claim 9 , wherein the highest ranked of the collected images is validated.
15. The system according to claim 13 , wherein the match probability metric includes selection criteria.
16. The system according to claim 15 , wherein the selection criteria is used to adjust future weighting.
17. The method according to claim 1 , wherein the architectural features include any of: windows, doors or signs.
18. The method according to claim 1 further comprising graphically highlighting regions identified with the differences based on a comparison of regions with metadata differing with regions of the highest ranked of the plurality of previously stored geospatial images.
19. A method of updating building model façade textures comprising:
receiving an oblique image captured by a smartphone, the received oblique image including one or more facades of a building;
processing regions of the received oblique image to remove façade obstructions;
processing regions of the received oblique image to identify architectural features of a selected facade within the regions as metadata associated with the received oblique image;
comparing the received oblique image to a plurality of previously stored images, the comparing including an analysis of the metadata;
ranking the comparing of the received oblique image and the plurality of previously stored images;
creating a pixel-to-pixel correlation between the received oblique image and a highest ranked one of the plurality of previously stored images;
storing the pixel-to-pixel correlated image in a database with metadata data derived during both the processing of the regions as well as with metadata associated with the highest ranked one of the plurality of previously stored images; and
when the pixel-to-pixel correlation indicates differences between the received oblique image and the highest ranked one of the plurality of previously stored images, updating the building model textures by retexturing at least one of the facades of a building model with the received oblique image.
20. The method according to claim 19 further comprising graphically highlighting regions identified with the differences based on a comparison of regions with metadata differing with regions of the highest ranked one of the plurality of previously stored images.