Absolute geospatial accuracy model for remote sensing without surveyed control points
Estimating absolute geospatial accuracy in input images without the use of surveyed control points is disclosed. For example, the absolute geospatial accuracy of a satellite images may be estimated without the use of control points (GCPs). The absolute geospatial accuracy of the input images may be estimated based on a statistical measure of relative accuracies between pairs of overlapping images. The estimation of the absolute geospatial accuracy may include determining a root mean square error of the relative accuracies between pairs of overlapping images. For example, the absolute geospatial accuracy of the input images may be estimated by determining a root mean square error of the shears of respective pairs of overlapping images. The estimated absolute geospatial accuracy may be used to curate GCPs, evaluate a digital elevation map, generate a heatmap, or determine whether the adjust the images until a target absolute geospatial accuracy is met.
1. A machine-implemented method, comprising:
accessing a set of input images, the set of input images comprising overlapping images;
determining relative accuracies between pairs of the overlapping images, including calculating a shear for each pair of overlapping images;
estimating absolute geospatial accuracy of the set of the input images, without absolute geospatial reference points, based on a statistical measure of the relative accuracies between the pairs of the overlapping images, including determining a statistic that is based on the shear for each pair of the overlapping images; and
filtering a set of ground control points by, for each ground control point:
determining an offset between the ground control point and a corresponding point in the set of the input images;
comparing the offset with the estimate of the absolute geospatial accuracy of the set of the input images; and
identifying the ground control point as having suspect accuracy responsive to the offset failing to meet an accuracy threshold.
2. The machine-implemented method of claim 1 , wherein estimating the absolute geospatial accuracy of the set of the input images, without absolute geospatial reference points, based on the statistical measure of the relative accuracies between the pairs of the overlapping images comprises:
determining a root mean square error of the relative accuracies between the pairs of the overlapping images.
3. The machine-implemented method of claim 2 , wherein estimating the absolute geospatial accuracy of the set of the input images, without absolute geospatial reference points, based on the statistical measure of the relative accuracies between the pairs of the overlapping images further comprises:
determining one or more terms to account for unmodeled or empirically known error; and
modifying the root mean square error based on the one or more terms.
4. The machine-implemented method of claim 1 , wherein determining the statistic that is based on the shear for each pair of overlapping images comprises:
determining a root mean square error based on the shear for each pair of the overlapping images.
5. The machine-implemented method of claim 1 , further comprising:
evaluating a digital elevation map (DEM) of a geographic region by:
estimating absolute geospatial accuracy of a set of Level 1B images of the geographic region based on a statistic that is based on a shear between pairs of overlapping Level 1B images in the set of Level 1B images;
estimating absolute geospatial accuracy of a set of orthorectified images of the geographic region based on a statistic that is based on a shear between pairs of overlapping orthorectified images in the set of orthorectified images; and
determining errors in the DEM based on differences between the estimated absolute geospatial accuracy of the set of orthorectified images and the estimated absolute geospatial accuracy of the set of Level 1B images.
6. The machine-implemented method of claim 5 , further comprising:
correcting the DEM based on the set of Level 1B images.
7. The machine-implemented method of claim 1 , further comprising:
estimating an absolute geospatial accuracy for each of a plurality of sets of input images of different geographic regions, including:
determining relative accuracies between pairs of overlapping images in each set of the plurality of sets of input images of different geographic regions, including calculating a shear for each pair of overlapping images in each set; and
estimating absolute geospatial accuracy of each set of the input images based on the shear for each pair of the overlapping images.
8. The machine-implemented method of claim 7 , further comprising:
generating an absolute geospatial accuracy heatmap of the different geographic regions based on the absolute geospatial accuracy of each respective geographic region.
9. The machine-implemented method of claim 1 , further comprising:
determining whether the estimate of the absolute geospatial accuracy of the set of input images meets an accuracy threshold; and
in response to determining that the estimate of the absolute geospatial accuracy of the set of input images does not meet the accuracy threshold, adjusting the accuracy of the set of input images until the accuracy threshold is met.
10. The machine-implemented method of claim 1 , further comprising:
calculating a confidence in the estimate of the absolute geospatial accuracy of the set of input images based on a number of images in the set of input images.
11. The machine-implemented method of claim 1 , wherein the set of input images comprise a mosaic.
12. The machine-implemented method of claim 1 , wherein the set of input images comprise a time series.
13. One or more non-transitory processor readable storage mediums that collectively store processor readable code that, when executed by one or more processors, cause the one or more processors to perform a method comprising:
accessing a set of Level 1B images of a geographic region, the set of Level 1B images comprising overlapping Level 1B images;
determining relative accuracies between pairs of the overlapping Level 1B images, including calculating a shear for each pair of overlapping Level 1B images;
estimating absolute geospatial accuracy of the set of the Level 1B images, without absolute geospatial reference points, based on a statistical measure of the relative accuracies between the pairs of the overlapping Level 1B images, including determining a statistic that is based on the shear for each pair of the overlapping Level 1B images;
accessing a set of orthorectified images of the geographic region, the set of orthorectified images comprising overlapping orthorectified images;
determining relative accuracies between pairs of the overlapping orthorectified images, including calculating a shear for each pair of overlapping orthorectified images;
estimating absolute geospatial accuracy of the set of the orthorectified images, without absolute geospatial reference points, based on a statistical measure of the relative accuracies between the pairs of the overlapping orthorectified images, including determining a statistic that is based on the shear for each pair of the overlapping orthorectified images; and
evaluating a digital elevation map (DEM) of the geographic region by determining errors in the DEM based on differences between the estimated absolute geospatial accuracy of the set of orthorectified images and the estimated absolute geospatial accuracy of the set of Level 1B images.
14. The one or more non-transitory processor readable storage mediums of claim 13 , wherein estimating the absolute geospatial accuracy of the set of the Level 1B images, without absolute geospatial reference points, based on the statistical measure of the relative accuracies between the pairs of the overlapping Level 1B images comprises:
determining a root mean square error of the relative accuracies between the pairs of the overlapping Level 1B images.
15. The one or more non-transitory processor readable storage mediums of claim 13 , wherein:
determining the statistic that is based on the shear for each pair of the overlapping Level 1B images comprises determining a root mean square error of the shear for each pair of the overlapping Level 1B images; and
determining the statistic that is based on the shear for each pair of the overlapping orthorectified images comprises determining a root mean square error of the shear for each pair of the overlapping orthorectified images.
16. The one or more non-transitory processor readable storage mediums of claim 13 , wherein the method further comprises:
filtering a set of ground control points by, for each ground control point:
determining an offset between the ground control point and a corresponding point in the set of the Level 1B images;
comparing the offset with the estimate of the absolute geospatial accuracy of the set of the Level 1B images; and
identifying the ground control point as having suspect accuracy responsive to the offset failing to meet an accuracy threshold.
17. The one or more non-transitory processor readable storage mediums of claim 13 , wherein the method further comprises:
estimating an absolute geospatial accuracy for each of a plurality of sets of Level 1B images of different geographic regions, wherein the absolute geospatial accuracy for each set is based on a shear for each pair of overlapping Level 1B images in the set; and
generating an absolute geospatial accuracy heatmap of the different geographic regions based on the absolute geospatial accuracy of each respective geographic region.
18. An apparatus, comprising:
one or more processors; and
one or more processor readable storage devices in communication with the one or more processors, the one or more processor readable storage devices store code for programming the one or more processors to:
access a set of input images, the set of input images comprising overlapping images;
determine relative accuracies between pairs of the overlapping images, including calculating a shear for each pair of overlapping images;
estimate absolute geospatial accuracy of the set of the input images, without absolute geospatial reference points, based on a statistical measure of the relative accuracies between the pairs of the overlapping images, including determine a statistic that is based on the shear for each pair of the overlapping images; and
filter a set of ground control points by, for each ground control point:
determine an offset between the ground control point and a corresponding point in the set of the input images;
compare the offset with the estimate of the absolute geospatial accuracy of the set of the input images; and
identify the ground control point as having suspect accuracy responsive to the offset failing to meet an accuracy threshold.
19. A machine-implemented method, comprising:
accessing a set of input images, the set of input images comprising overlapping images;
determining relative accuracies between pairs of the overlapping images, including calculating a shear for each pair of overlapping images;
estimating absolute geospatial accuracy of the set of the input images, without absolute geospatial reference points, based on a statistical measure of the relative accuracies between the pairs of the overlapping images, including determining a statistic that is based on the shear for each pair of the overlapping images; and
evaluating accuracy of a set of ground control points based on the estimate of the absolute geospatial accuracy of the set of the input images.
20. The machine-implemented method of claim 19 , wherein evaluating the accuracy of the set of ground control points based on the estimate of the absolute geospatial accuracy of the set of the input images further comprises:
identifying ground control points in the set of ground control points as having suspect accuracy based on the estimate of the absolute geospatial accuracy of the set of the input images.