SYSTEMS AND METHODS FOR DIGITAL SURFACE MODEL RECONSTRUCTION FROM IMAGES USING ARTIFICIAL INTELLIGENCE
Systems and methods for creating digital surface models (DSMs) are disclosed, including a method comprising generating, with machine learning algorithm(s), a candidate DSM of a first geographic area with first image(s) from a set of first images depicting a first characteristic, the candidate DSM having voxels identifying a location within the first geographic area and having an elevation value; comparing elevation values for voxels of the candidate DSM to corresponding elevation values for voxels of a predetermined DSM, created with a set of second images of the first geographic area having a second characteristic including features beyond that provided with the set of first images, to determine error; adjusting, via back-propagation, the machine learning algorithm(s) based on the determined error; and generating with the trained machine learning algorithm(s), a DSM using a set of third images depicting a second geographic area.
1 . A method for creating digital surface models, comprising:
a. generating, with one or more machine learning algorithms, a candidate digital surface model of a portion of a first geographic area with one or more first images from a set of first images, the candidate digital surface model having a plurality of voxels with at least some of the voxels identifying a location within the first geographic area and having an elevation value, the set of first images depicting at least a portion of the first geographic area with a first characteristic;
b. comparing elevation values for voxels of the candidate Digital Surface Model to corresponding elevation values for voxels of a predetermined Digital Surface Model of the first geographic area to determine an error for the voxels of the candidate Digital Surface Model, the predetermined Digital Surface Model created with a set of second images of the first geographic area having a second characteristic including features beyond that provided with the set of first images having the first characteristic;
c. adjusting, via back-propagation, the one or more machine learning algorithms based on the determined error for the voxels of the candidate Digital Surface Model; and
d. repeating a. b. and c. until the determined errors for the voxels is below a predetermined threshold indicating trained machine learned algorithms; and
e. generating with the trained machine learning algorithms, a new Digital Surface Model using a set of third images depicting a second geographic area.
2 . The method for creating digital surface models of claim 1 , wherein the first geographic area covers a different geographic region than the second geographic area.
3 . The method for creating digital surface models of claim 1 , wherein the first geographic area and the second geographic area have similar features.
4 . The method for creating digital surface models of claim 1 , wherein the first characteristic is deciduous trees without leaves, and the second characteristic including features beyond that provided with the set of first images having the first characteristic is deciduous trees with leaves.
5 . The method for creating digital surface models of claim 1 , wherein the first characteristic is one or more objects having an obstruction, and the second characteristic including features beyond that provided with the set of first images having the first characteristic is one or more objects without the obstruction.
6 . The method for creating digital surface models of claim 1 , wherein one or more of the first images, the second images, and the third images are geo-referenced images having one or more pixels having associated geolocation data.
7 . The method for creating digital surface models of claim 6 , further comprising:
incorporating terrain elevation into one or more of the candidate Digital Surface Model, the predetermined Digital Surface Model, and the new Digital Surface Model.
8 . The method for creating digital surface models of claim 7 , wherein incorporating terrain elevation comprises utilizing the geolocation data of the geo-reference images to associate corresponding terrain elevation data with the one or more pixels.
9 . A system for creating digital surface models, comprising:
one or more 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:
a. generate, with one or more machine learning algorithms, a candidate Digital Surface Model of a portion of a first geographic area with one or more first images from a set of first images, the candidate Digital Surface Model having a plurality of voxels with at least some of the voxels identifying a location within the first geographic area and having an elevation value, the set of first images depicting at least a portion of the first geographic area with a first characteristic;
b. compare elevation values for voxels of the candidate Digital Surface Model to corresponding elevation values for voxels of a predetermined Digital Surface Model of the first geographic area to determine an error for the voxels of the candidate Digital Surface Model, the predetermined Digital Surface Model created with a set of second images of the first geographic area having a second characteristic including features beyond that provided with the set of first images having the first characteristic;
c. adjust, via back-propagation, the one or more machine learning algorithms based on the determined error for the voxels of the candidate Digital Surface Model; and
d. repeat a. b. and c. until the determined errors for the voxels is below a predetermined threshold indicating trained machine learned algorithms; and
e. generate with the trained machine learning algorithms, a new Digital Surface Model using a set of third images depicting a second geographic area.
10 . The system for creating digital surface models of claim 9 , wherein the first geographic area covers a different geographic region than the second geographic area.
11 . The system for creating digital surface models of claim 9 , wherein the first geographic area and the second geographic area have similar features.
12 . The system for creating digital surface models of claim 9 , wherein the first characteristic is deciduous trees without leaves, and the second characteristic including features beyond that provided with the set of first images having the first characteristic is deciduous trees with leaves.
13 . The system for creating digital surface models of claim 9 , wherein the first characteristic is one or more objects having an obstruction, and the second characteristic including features beyond that provided with the set of first images having the first characteristic is one or more objects without the obstruction.
14 . The system for creating digital surface models of claim 9 , wherein one or more of the first images, the second images, and the third images are geo-referenced images having one or more pixels having associated geolocation data.
15 . The system for creating digital surface models of claim 9 , the one or more 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:
incorporate terrain elevation into one or more of the candidate Digital Surface Model, the predetermined Digital Surface Model, and the new Digital Surface Model.
16 . The system for creating digital surface models of claim 15 , wherein one or more of the first images, the second images, and the third images are geo-referenced images having one or more pixels having associated geolocation data, and wherein incorporating terrain elevation comprises utilizing the geolocation data of the geo-referenced images to associate corresponding terrain elevation data with the one or more pixels.
17 . The system for creating digital surface models of claim 12 , the one or more 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:
determine solar access values for a roof based on the new Digital Surface Model, and calculate a ray between a sun position and the roof as affected by the new Digital Surface Model in relation to a path of the ray.
18 . The system for creating digital surface models of claim 9 , wherein the machine learning algorithms comprise Generative Adversarial Networks (GANs).
19 . A system for creating digital surface models, comprising:
one or more 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:
create a digital surface model of a desired geographic area from one or more desired digital images depicting the desired geographic area, the digital surface model depicting objects having a first characteristic including features beyond that depicted in the one or more desired digital images, by utilizing trained machine learning algorithms, the trained machine algorithms having been trained by:
a. generating, with one or more original machine learning algorithms, a candidate Digital Surface Model of a portion of a first geographic area with one or more first images from a set of first images, the candidate Digital Surface Model having a plurality of voxels with at least some of the voxels identifying a location within the first geographic area and having an elevation value, the set of first images depicting at least a portion of the first geographic area with the first characteristic;
b. determining error for the voxels of the candidate Digital Surface Model by comparing elevation values for voxels of the candidate Digital Surface Model to corresponding elevation values for voxels of a predetermined Digital Surface Model of the first geographic area, the predetermined Digital Surface Model created with a set of second images of the first geographic area having a second characteristic including features beyond that provided with the set of first images having the first characteristic;
c. adjusting, via back-propagation, the one or more machine learning algorithms based on the determined error for the voxel of the candidate Digital Surface Model; and
d. repeating a. b. and c. until the determined errors for the voxels is below a predetermined threshold indicating trained machine learned algorithms.
20 . The system for creating digital surface models of claim 19 , wherein the one or more desired digital images are not required to be part of stereo image pairs.