In-scene determination of aerosol parameters from imagery

Non-transitory computer-readable storage media, storing instructions that, when read by the computer, cause the computer to perform a method, the method comprising:

• receiving multispectral or hyperspectral computer imagery of the scene, from which radiance measurements in at least two spectral bands can be obtained;
• receiving a delineation within the imagery of at least one calibration target made from a known material;
• determining from the imagery a radiance measurement for each target of the at least one calibration target in multiple spectral bands including each of at least two spectral bands having distinct wavelength ranges;
• determining from the imagery a single radiance calibration reference for a background common to each target the at least one calibration target;
• receiving for the material of which each of the at least one calibration targets, a predetermined reflectance in the multiple spectral bands, the single calibration reference excluding non-background radiance measurements from the at least one calibration target;
• for each of a predefined plurality of aerosol types, each aerosol being defined by a plurality of properties, fitting a background reflectance of the scene to radiance measurements of the at least one calibration target and the single radiance calibration reference for the background, using at least the properties of the aerosol and the spectral reflectance of the at least one calibration target in each of the multiple spectral bands;
• and
• selecting the predefined aerosol that results in the best fit, the properties of the predefined aerosol being estimates for the properties of the aerosol present in the atmosphere of the scene.

An apparatus for estimating properties of aerosol present in the atmosphere of a scene from imagery taken of the scene, comprising:

• means for storing multispectral or hyperspectral imagery of the scene, from which radiance measurements in at least two spectral bands can be obtained;
• means for determining, in response to delineation within the imagery of at least one calibration target made from a known material and a background common to each of the at least one calibration target, from the imagery a radiance measurement for each target of the at least one calibration target in multiple spectral bands including each of at least two spectral bands having distinct wavelength ranges, and from the imagery a single radiance calibration reference for the background, the single calibration reference excluding non-background radiance measurements from the at least one calibration target;
• means for fitting, for each of a predefined plurality of aerosol types, each aerosol being defined by a plurality of properties a background reflectance of the scene to radiance measurements of the at least one calibration target and the single radiance calibration reference for the background, using at least the properties of the aerosol and, for each target of the at least one calibration target in each of the multiple spectral bands, a previously determined spectral reflectance for the material of which it is made;
• and
• means for selecting the predefined aerosol that results in the best fit, the properties of the predefined aerosol being estimates for the properties of the aerosol present in the atmosphere of the scene.

A computer implemented method for estimating properties of aerosol present in the atmosphere of a scene from imagery taken of the scene, comprising:

• receiving with a specially programmed computer imagery of the scene, from which radiance measurements in at least two spectral bands can be obtained;
• receiving with the specially programmed computer a delineation within the imagery of at least one calibration target made from a known material;
• determining from the imagery a radiance measurement for each of the at least one calibration targets in each of at least two spectral bands;
• determining from the imagery a radiance measure for a background common to all of the at least one calibration target, but excluding the at least one calibration target;
• for each of a predefined plurality of aerosol types, each aerosol being defined by a plurality of properties, fitting with a fitting process executing on the specially programmed computer a background reflectance of the scene to the measured radiances of the at least one calibration target and the background, using at least the properties of the aerosol and a predetermined spectral reflectance of the at least one calibration target;
• selecting the predefined aerosol that results in the best fit, the properties of the predefined aerosol being estimates for the properties of the aerosol present in the atmosphere of the scene;
• wherein the fitting process is comprised of a least square fitting process, and wherein the least square fitting process is comprised of a chi-square fitting process;
• wherein the chi-square fitting process for a given, predefined aerosol type, comprises,
• determining an initial estimate for a state space defined by a gray body equivalent background reflectance in each of the at least two bands, and a visibility for the scene, the background reflectance being determined by solving for the background reflectance by forward model functions mapping the background reflectance and visibility to measured radiances of the at least one calibration target and the background and the spectral reflectance of the at least one calibration target;
• and
• minimizing a chi-square penalty function for the state space, the chi-square penalty function being defined in terms of a difference between a radiance for each of the calibration targets in each of the at least two bands, as determined by the forward model functions, and the measured radiances for the at least one calibration target in each of the at least two bands
• wherein the forward model functions are comprised of Nb·(Nr+1) functions denoted Li,j:RNb+1→R for i=0, K, Nr and j=0, K, Nb−1, wherein Li,j(ρj,ν;ρir) is defined by:
• L
• i
• ,
• j
• ⁡
• (
• ρ
• j
• ,
• v
• )
• =
• {
• ∫
• [
• L
• AS
• +
• ρ
• i
• r
• 1
• -
• ρ
• j
• ⁢
• S
• ⁢
• L
• SR
• +
• ρ
• j
• 1
• -
• ρ
• j
• ⁢
• S
• ⁢
• L
• BSR
• ]
• ⁢
• R
• j
• ⁡
• (
• λ
• )
• ⁢
• ⅆ
• λ
• i
• =
• 0
• ,
• K
• ,
• N
• r
• -
• 1
• ∫
• [
• L
• AS
• +
• ρ
• j
• 1
• -
• ρ
• j
• ⁢
• S
• ⁢
• (
• L
• SR
• +
• L
• BSR
• )
• ]
• ⁢
• R
• j
• ⁡
• (
• λ
• )
• ⁢
• ⅆ
• λ
• i
• =
• N
• r
• wherein Nr≧1 is a number of calibration targets within the imagery and Nb is the number of bands of a sensor acquiring the imagery;
• and
• wherein LSR=LDSR+LSSR and LBSR=LBDSR+LBSSR;
• LAS(λ) being solar photons that are scattered into the sensor's field of view via single or multiple scatter events within the atmosphere without ever reaching the target or background;
• LDSR(λ) being solar photons that pass directly through the atmosphere to a 100% reflectance Lambertian target, reflect off the target, and propagate directly through the atmosphere to a sensor that acquires the imagery;
• LSSR(λ) being Solar photons that are scattered by the atmosphere onto a 100% reflectance Lambertian target, reflect off the target, and propagate directly through the atmosphere to the sensor;
• LBDSR(λ) being solar photons that are directly transmitted to a 100% reflectance Lambertian background, reflect off the background once, and then scatter into the field of view of the sensor;
• LBSSR(λ) being solar photons that are scattered by the atmosphere at least once, reflect off a 100% reflectance Lambertian background once, and then scatter into the field of view of the sensor;
• S(λ) being a spherical albedo of the bottom of the atmosphere, which can be thought of as a reflection coefficient of the atmosphere;
• and Rj(λ) being the relative spectral response function of the sensor in band j.