Method and system for aerial image processing
A method for aerial image processing preferably includes receiving ground control point (GCP) information, receiving a set of images, determining spatial information associated with the images, and determining GCP image locations; and can additionally or alternatively include determining aerial map information, training an automated GCP detector, and/or any other suitable elements. A system for aerial image processing preferably includes one or more computing devices, and can additionally or alternatively include one or more GCPs, aerial vehicles, and/or any other suitable elements.
1. A method for ground control point (GCP) mapping, comprising:
receiving a GCP dataset associated with a mapping region, the GCP dataset comprising a geographic location associated with a GCP;
receiving a set of images associated with the mapping region, each image of the set associated with a respective camera pose estimate, wherein each camera pose estimate comprises a camera position estimate and a camera orientation estimate;
selecting a plurality of images of the set, wherein, for each image of the plurality, selecting the image comprises predicting that the image depicts the GCP based on the geographic location and the respective camera pose estimate associated with the image;
for each image of the plurality, automatically determining a respective GCP image location within the image using a computer vision technique;
after automatically determining the GCP image locations, determining, based on the GCP image locations and the camera pose estimates, that a first GCP image location is incorrect, comprising:
for each GCP image location, determining a respective implied GCP position based on the camera pose estimates;
based on the implied GCP positions, determining a consolidated implied GCP position; and
determining that a first implied GCP position, associated with the first GCP image location, is not within a threshold distance of the consolidated implied GCP position:
after determining that the first GCP image location is incorrect, refining the camera pose estimates based on the geographic location and the GCP image locations, wherein refining the camera pose estimates is not performed based on the first GCP image location; and
after refining the camera pose estimates, generating a representation of the mapping region based on the set of images and the camera pose estimates.
2. The method of claim 1 , wherein the representation of the mapping region comprises an orthomosaic map of the mapping region.
3. The method of claim 1 , further comprising, before further refining the camera pose estimates based on the geographic location and the GCP image locations:
displaying a representation of the respective GCP image location to a user; and
after displaying the representation of the respective GCP image location to the user, receiving a confirmation input from the user indicative of GCP representation presence at the respective GCP image location.
4. The method of claim 1 , wherein the representation of the mapping region comprises an elevation map of the mapping region.
5. The method of claim 1 , further comprising, before automatically determining the respective GCP image locations, refining the camera pose estimates based on the set of images.
6. The method of claim 5 , wherein refining the camera pose estimates based on the set of images comprises:
based on a first image of the set of images, determining a set of features, each feature of the set associated with a respective first image subset of the first image, wherein the respective first image subset depicts the feature;
determining a set of matches, comprising, for each feature of the set, determining a respective second image subset of a second image of the set of images, wherein the respective second image subset depicts the feature;
based on the set of matches, determining a set of feature positions, comprising, for each feature of the set, determining a respective feature position of the feature within the mapping region; and
based on the set of feature positions, determining refined camera pose estimates associated with the first and second images.
7. The method of claim 6 , wherein refining the camera pose estimates based on the geographic location and the GCP image locations comprises:
determining a refined set of feature positions based on the geographic location and the GCP image locations; and
based on the refined set of feature positions, determining further refined camera pose estimates associated with the first and second images.
8. The method of claim 6 , wherein determining the set of feature positions comprises:
based on the set of matches, determining a set of estimated positions; and
performing a bundle adjustment to refine the set of estimated positions.
9. The method of claim 5 , wherein refining the camera pose estimates based on the set of images comprises:
based on a first image of the set of images, determining a set of features, each feature of the set associated with a respective first image subset of the first image, wherein the respective first image subset depicts the feature;
determining a set of matches, comprising, for each feature of the set, determining a respective second image subset of a second image of the set of images, wherein the respective second image subset depicts the feature;
based on the set of matches, determining a set of feature positions, comprising, for each feature of the set, determining a respective feature position of the feature within the mapping region; and
based on the set of feature positions, determining refined camera pose estimates associated with the first and second images.
10. The method of claim 1 , wherein determining the consolidated implied GCP position comprises:
partitioning the implied GCP positions into a plurality of subsets using a clustering technique; and
selecting a subset of the plurality as a primary subset, wherein:
the primary subset has a cardinality greater than a cardinality of every other subset of the plurality; and
the primary subset does not comprise the first implied GCP position; and
determining a central position based on the primary subset, wherein:
each implied GCP position of the primary subset is within the threshold distance of the central position; and
the consolidated implied GCP position is the central position.
11. The method of claim 1 , wherein automatically determining the respective GCP image location within the image using the computer vision technique comprises using a convolutional neural network trained to recognize depictions of GCPs.
12. The method of claim 1 , further comprising, before refining the camera pose estimates based on the geographic location and the GCP image locations, for each image of the plurality:
displaying a representation of the respective GCP image location to a user; and
after displaying the representation of the respective GCP image location to the user, receiving a confirmation input from the user indicative of GCP representation presence at the respective GCP image location.