Abstract
Test sites on the surface of the Earth, used for vicarious radiometric calibration, generally consist of an extended uniform surface filling a large area of the sensor’s focal plane. The ability to sample the site with a large number of detectors establishes a mean response to the site’s predicted at–sensor reference radiance with high precision and averages out the blurring effects induced by the imaging system. The procedure derives radiometric gain coefficients with uncertainties that can be maintained through the image processing chain to higher level products but only for extended uniform scenes. However, many targets of interest are small, typically subpixel to a few pixels in diameter. As a result, the spatial performance of the sensor and any pixel resampling performed in the processing chain can become a significant error contributor to the radiometric knowledge of small targets. This leads to several factors to be accounted for in the accuracy and traceability bookkeeping specific to small targets. First, is the non-uniform energy-on-detector sensitivity. In this effect, the shape of the small target image profile projected on the focal plane detectors stands in contrast to the scene background in which a single detector may ultimately respond differently to illumination gradients not present when under calibration. Second is sensor blurring. With sensor blurring the energy originating geometrically within the projected ground location of a single detector, containing the small target, becomes recorded by adjacent detectors outside the target area because of the system point-spread function (PSF). This induces a probabilistic uncertainty in the location of the signal’s origin and thus the target radiance without some a priori knowledge of the targets shape and size. And third are the effects of post processing. In post processing, band-to-band registration and geometric calibration apply resampling methods that ultimately introduce a radiometric bias changing the contrast between the small target and its background specific to the routine used to redefine individual pixel digital values, size and orientation. As a result, a goal to develop a comprehensive radiometric error propagation analysis process for small targets can benefit by introducing ground references that reveal and validate these artifacts within image data collected under operational conditions. In this presentation, the vicarious calibration technique known as the Specular Array Radiometric Calibration (SPARC) method is demonstrated to provide a highly flexible and robust approach for characterizing the performance of a sensor imaging small targets. SPARC delivers small reference targets, made from a collection of facets consisting of convex mirrors, which can be built to any size, shape, brightness or spectral composition with accurate radiometric traceability to the solar radiometric scale. Examples of assessing the spatial, radiometric, spectral and geometric performance of several civil and commercial sensors against small SPARC targets will be presented.
Using Vicarious Calibration to Evaluate Small Target Radiometry
Test sites on the surface of the Earth, used for vicarious radiometric calibration, generally consist of an extended uniform surface filling a large area of the sensor’s focal plane. The ability to sample the site with a large number of detectors establishes a mean response to the site’s predicted at–sensor reference radiance with high precision and averages out the blurring effects induced by the imaging system. The procedure derives radiometric gain coefficients with uncertainties that can be maintained through the image processing chain to higher level products but only for extended uniform scenes. However, many targets of interest are small, typically subpixel to a few pixels in diameter. As a result, the spatial performance of the sensor and any pixel resampling performed in the processing chain can become a significant error contributor to the radiometric knowledge of small targets. This leads to several factors to be accounted for in the accuracy and traceability bookkeeping specific to small targets. First, is the non-uniform energy-on-detector sensitivity. In this effect, the shape of the small target image profile projected on the focal plane detectors stands in contrast to the scene background in which a single detector may ultimately respond differently to illumination gradients not present when under calibration. Second is sensor blurring. With sensor blurring the energy originating geometrically within the projected ground location of a single detector, containing the small target, becomes recorded by adjacent detectors outside the target area because of the system point-spread function (PSF). This induces a probabilistic uncertainty in the location of the signal’s origin and thus the target radiance without some a priori knowledge of the targets shape and size. And third are the effects of post processing. In post processing, band-to-band registration and geometric calibration apply resampling methods that ultimately introduce a radiometric bias changing the contrast between the small target and its background specific to the routine used to redefine individual pixel digital values, size and orientation. As a result, a goal to develop a comprehensive radiometric error propagation analysis process for small targets can benefit by introducing ground references that reveal and validate these artifacts within image data collected under operational conditions. In this presentation, the vicarious calibration technique known as the Specular Array Radiometric Calibration (SPARC) method is demonstrated to provide a highly flexible and robust approach for characterizing the performance of a sensor imaging small targets. SPARC delivers small reference targets, made from a collection of facets consisting of convex mirrors, which can be built to any size, shape, brightness or spectral composition with accurate radiometric traceability to the solar radiometric scale. Examples of assessing the spatial, radiometric, spectral and geometric performance of several civil and commercial sensors against small SPARC targets will be presented.