Remote sensing imagery accuracy analysis method and apparatus
View Patent ↗A method of enhancing a resolution of an image by fusing images includes applying a principal component analysis to a multispectral image to obtain a plurality of principal components, and replacing a first component in the plurality of principal components by a panchromatic image. The method further includes resampling remaining principal components to a resolution of the panchromatic image, and applying an inverse principal analysis to the panchromatic image and the remaining principal components to obtain a fused image of the panchromatic image and the multispectral image.
1. A method for analyzing and estimating horizontal accuracy in georeferenced images, the method comprising:
inputting into a computer-based device remotely acquired georeferenced images, photo-identified locations from georeferenced images, and real world survey locations of objects on the surface of the earth in x and y directions;
calculating using a computer-based device an offset error between georeferenced image locations and real world survey locations of objects on the surface of the earth in both x and y directions;
deriving a root mean square error from the calculated offset error in the x and y directions;
deriving a circular error probability from the derived root mean square error in the x and y directions; and
computing using a computer-based device the horizontal accuracy between georeferenced image locations and real world survey locations of objects on the surface of the earth by using the root mean square error in the x and y directions and circular error probability for geolocation mapping of object locations on the surface of the earth.
2. The method of claim 1 , wherein the analyzing and estimating of the horizontal accuracy further comprises calculating a set of statistical and mathematical accuracy metrics which further comprises:
determining circular error probability wherein 90% of well-defined points in an image or map should fall within a certain radial distance (CE 90 ), wherein
CE 90 =2.1460σ c ;
determining circular error probability wherein 95% of well-defined points in an image or map should fall within a certain radial distance (CE 95 ), wherein
CE 95 =2.4477σ c ;
and wherein σ x =σ y =σ c and the root mean square error in X and Y are calculated as
σ
x
=
∑
(
x
image
-
x
realworld
)
2
n
σ
y
=
∑
(
y
image
-
y
realworld
)
2
n
and x image and x realworld and y image and y realworld are coordinates from photo-identifiable points in the georeferenced image and survey points from the real world and n is the number of control points;
estimating σ c , where σ x and σ y are not equal, by a linear combination of σ x and σ y ; and
estimating σ c , where the
σ
min
σ
max
ratio is between 0.6 and 1.0, by a linear combination of σ x and σ y where
σ c =0.5222σ min +0.4778σ max
and where the
σ
min
σ
max
ratio is less than 0.6, interpolated coefficients are utilized to produce varying linear combinations of σ x and σ y ,
where σ min is the minimum value between σ x and σ y and σ max is the maximum value between σ x and σ y .