Geospatial object geometry extraction from imagery
Apparatuses, systems, methods, and medium are disclosed for precise geospatial structure geometry extraction from multi-view imagery, including a non-transitory computer readable medium storing computer executable code that when executed by a processor cause the processor to: receive an image of a structure having an outline, the image having pixels with first pixel values depicting the structure and second pixel values outside of the structure depicting a background of a geographic area surrounding the structure, and image metadata including first geolocation data; and generate a synthetic shape image of the structure from the image using a machine learning algorithm, the synthetic shape image including pixels having pixel values forming a synthetic shape of the outline, the synthetic shape image having second geolocation data derived from the first geolocation data.
1. A non-transitory computer readable medium storing computer executable code that when executed by one or more computer processors causes the one or more computer processors to:
receive an image of a structure having an outline, the image having pixels with first pixel values depicting the structure and second pixel values outside of the structure depicting a background of a geographic area surrounding the structure, and image metadata including first geolocation data;
generate a synthetic shape image of the structure from the image using a machine learning algorithm, the synthetic shape image including pixels having pixel values forming a synthetic shape of the outline, the synthetic shape image having second geolocation data derived from the first geolocation data, wherein the synthetic shape of the outline is a vector outline of the structure;
map the vector outline onto the image of the structure, based at least in part on the first and second geolocation data; and
change the second pixel values of the image so as to not depict the background of the geographic area surrounding the structure.
2. The non-transitory computer readable medium of claim 1 , wherein the computer executable code that when executed by the one or more computer processors further causes the one or more computer processors to measure the vector outline of the structure from the image.
3. The non-transitory computer readable medium of claim 1 , wherein the image has a pixel resolution less than 10 inches per pixel.
4. The non-transitory computer readable medium of claim 1 , wherein the image has a pixel resolution between 10 inches per pixel and 0.1 inches per pixel.
5. The non-transitory computer readable medium of claim 1 , wherein the machine learning algorithm is a first machine learning algorithm, and wherein the first machine learning algorithm is a component of a generator of a generative adversarial network, the generative adversarial network further comprising a discriminator having a second machine learning algorithm, the generator receiving the image of the structure and generating the synthetic shape image.
6. The non-transitory computer readable medium of claim 5 , wherein the generative adversarial network has been trained with truth pairs with each truth pair including a truth image and a truth shape image.
7. The non-transitory computer readable medium of claim 6 , wherein the truth image and the truth shape image have a same pixel resolution.
8. The non-transitory computer readable medium of claim 6 , wherein the truth image includes third geolocation data and the truth shape image includes fourth geolocation data, the fourth geolocation data being derived from the third geolocation data.
9. The non-transitory computer readable medium of claim 1 , wherein the image is a nadir image.
10. A method, comprising:
supplying a plurality of truth pairs to a first machine learning algorithm and a second machine learning algorithm, each of the truth pairs including a truth image and a truth shape image, the truth image having first pixel values depicting a structure and second pixel values depicting a background of a geographic area surrounding the structure, the structure having an outline, the truth shape image having third pixel values indicative of a truth shape indicating the outline of the structure, wherein the truth image and the truth shape image have a same pixel resolution;
generating a synthetic shape image of the structure from the truth image using the first machine learning algorithm, the synthetic shape image including pixels having fourth pixel values forming a synthetic shape of the outline of the structure;
passing the synthetic shape image of the structure from the first machine learning algorithm to the second machine learning algorithm;
comparing the synthetic shape against a truth shape from the truth shape image; and
providing feedback from the second machine learning algorithm to the first machine learning algorithm to train the first machine learning algorithm to minimize any differences in the synthetic shape and the truth shape.
11. The method of claim 10 , wherein the first machine learning algorithm is a component of a generator of a generative adversarial network, and the second machine learning algorithm is a component of a discriminator.
12. A method, comprising:
receiving, with one or more computer processors, an image of a structure having an outline, the image having pixels with first pixel values depicting the structure and second pixel values outside of the structure depicting a background of a geographic area surrounding the structure, and image metadata including first geolocation data;
generating a synthetic shape image of the structure from the image using a machine learning algorithm, the synthetic shape image including pixels having pixel values forming a synthetic shape of the outline, the synthetic shape image having second geolocation data derived from the first geolocation data, wherein the synthetic shape of the outline is a vector outline of the structure;
mapping, with the one or more computer processors, the vector outline onto the image of the structure, based at least in part on the first and second geolocation data; and
changing the second pixel values of the image so as to not depict the background of the geographic area surrounding the structure.
13. The method of claim 12 , further comprising measuring, with the one or more computer processors, the vector outline of the structure from the image.
14. The method of claim 12 , wherein the image has a pixel resolution less than 10 inches per pixel.
15. The method of claim 12 , wherein the machine learning algorithm is a first machine learning algorithm, and wherein the first machine learning algorithm is a component of a generator of a generative adversarial network, the generative adversarial network further comprising a discriminator having a second machine learning algorithm, the generator receiving the image of the structure and generating the synthetic shape image.
16. The method of claim 15 , wherein the generative adversarial network has been trained with truth pairs with each truth pair including a truth image and a truth shape image.
17. The method of claim 16 , wherein the truth image includes third geolocation data and the truth shape image includes fourth geolocation data, the fourth geolocation data being derived from the third geolocation data.