Systems and methods for relating images to each other by determining transforms without using image acquisition metadata
View Patent ↗Systems and methods for relating images to each other by determining transform functions between the images and the three-dimensional world coordinate system of the object depicted in the image, without using image acquisition metadata are disclosed. Points of interest are independently selected on each image. An initial transform function is applied to transform the points in the plane of one image to the plane of the other image. A Gaussian Mixture Model is then iteratively applied to the points to determine a best match, which then provides adjustments to the argument values of the transform function which is again applied to the points of interest on each image. This process repeats until the argument values converge to provide the resulting transform function. The resulting function can then be used to identify objects in the images in three dimensional space.
1. A computer-implemented method comprising:
receiving a first and a second image of an object, the images taken from different locations;
identifying the location of a set of points from the first image, the set of points indicating features of interest relative to a first image plane;
identifying the location of a set of points from the second image, the set of points indicating features of interest relative to a second image plane;
identifying argument values initially for a transform function, the transform function configured to map the location of the set of points relative to the first image plane to the location of the points relative to the second image plane;
iteratively refining the transform function arguments; and
outputting the transform function arguments.
2. The method of claim 1 wherein refining the transform function arguments further comprises:
applying the transform function with the argument values to map the location of a subset of the set of points from the first image in the first image plane to a location on the second image plane,
performing a Gaussian Mixture Model to fit the location of the subset of the set of points from the first image on the second image plane with the location of the set of points from the second image on the second image plane,
evaluating the fit of the Gaussian Mixture Model, and
updating the arguments of the transform function.
3. The method of claim 1 wherein the set of points indicating features of interest represents one or more of: corners of the roof and points of intersection of planar sections of the roof.
4. The method of claim 1 wherein the transform function is further configured to map the location of the set of points relative to the real-world three dimensional coordinate system of the object.
5. The method of claim 4 further comprising:
determining the height of one point in the set of points in the real-world three dimensional coordinate system of the object.
6. The method of claim 2 wherein evaluating the fit of the Gaussian Mixture Model further comprises:
creating a first graphical representation of the subset of the set of points from the first image on the first image plane;
creating a second graphical representation of the set of points from the second image; and
comprising the first graphical representation to the second graphical representation.
7. A computation machine comprising:
a computer processor; and
a memory coupled to the computer processor, wherein the memory has computer-executable instructions stored thereon that when executed by the computer processor, configure the computer processor to:
receive a first and a second image of an object, the images taken from different locations;
identify the location of a set of points from the first image, the set of points indicating features of interest relative to a first image plane;
identify the location of a set of points from the second image, the set of points indicating features of interest relative to a second image plane;
identify argument values initially for a transform function, the transform function configured to map the location of the set of points relative to the first image plane to the location of the points relative to the second image plane;
iteratively refine the transform function arguments; and
output the transform function arguments.
8. The computation machine of claim 7 wherein the set of points indicating features of interest represent one or more of: corners of the roof and points of intersection of planar sections of the roof.
9. A non-transitory computer readable medium having computer executable instructions thereon that, when executed, cause a computer processor to perform the following:
receiving a first and a second image of an object, the images taken from different locations;
identifying the location of a set of points from the first image, the set of points indicating features of interest relative to a first image plane;
identifying the location of a set of points from the second image, the set of points indicating features of interest relative to a second image plane;
identifying argument values initially for a transform function, the transform function configured to map the location of the set of points relative to the first image plane to the location of the points relative to the second image plane;
iteratively refining the transform function arguments; and
outputting the transform function arguments.
10. A method comprising:
receiving a first and a second image of an object, the images taken from different locations;
identifying the location of a set of points from the first image, the set of points indicating features of interest relative to a first image plane;
identifying the location of a set of points from the second image, the set of points indicating features of interest relative to a second image plane;
identifying argument values initially for a transform function, the transform function configured to map the location of the set of points relative to the first image plane to the location of the points relative to the second image plane;
iteratively refining the transform function arguments by carrying out the following steps:
applying the transform function with the argument values to map the location of a subset of the set of points from the first image in the first image plane to a location on the second image plane,
performing a Gaussian Mixture Model to fit the location of the subset of the set of points from the first image on the second image plane with the location of the set of points from the second image on the second image plane,
evaluating the fit of the Gaussian Mixture Model, and
updating the arguments of the transform function; and
outputting the updated arguments of the transform function.
11. The computer-implemented method of claim 10 , wherein evaluating the fit of the Gaussian Mixture Model further comprises:
creating a first graphical representation of the subset of the set of points from the first image on the first image plane;
creating a second graphical representation of the set of points from the second image; and
comparing the first graphical representation to the second graphical representation.
12. The computation machine of claim 7 , wherein the transform function is further configured to map the location of the set of points relative to the real-world three dimensional coordinate system of the object.
13. The computation machine of claim 12 , further comprising:
determine the height of one point in the set of points in the real-world three dimensional coordinate system of the object.
14. The non-transitory computer readable medium of claim 9 , wherein the set of points indicating features of interest represent one or more of: corners of the roof and points of intersection of planar sections of the roof.
15. The non-transitory computer readable medium of claim 9 , wherein the transform function is further configured to map the location of the set of points relative to the real-world three dimensional coordinate system of the object.
16. The non-transitory computer readable medium of claim 15 further comprising:
determining the height of one point in the set of points in the real-world three dimensional coordinate system of the object.