High spatial resolution imaging spectroscopy data is affected by various shadowing influences: be it the shadows of buildings, within forests, at forest borders, by single trees, by terrain, or by clouds. The radiometric performance of modern imaging spectroscopy systems provides a dynamic range which is large enough to retrieve useful spectral surface information even within full cast shadows. Thus, atmospheric correction routines should provide a capability to account for all above-mentioned types of cast shadows. The use of accurate surface models for cast shadow ray tracing does not lead to useful results in many cases as the surface representation with respect to the radiometry is never accurate enough. Therefore, an image-based method applicable to a broad variety of image data is searched. A method for cast shadow detection and correction has been implemented and included in the workflow of the ATCOR atmospheric compensation package. The shadows are retrieved from calibrated at-sensor radiance imagery relying on the fact that all areas in shade are illuminated by a large proportion of diffuse irradiance. The diffuse irradiance is caused by scattering in the atmosphere and thus exhibits very specific spectral characteristics compared to the direct irradiance. Specifically, the relative signal in the blue is significantly higher in areas affected by cast shadow than in directly illuminated regions. Specific indices have been defined to quantify the shadow fraction on a per-pixel basis by exploiting the spectral properties of illumination. This strategy results in a continuous quantification of the shadow field. The output is used directly in the atmospheric/topographic correction on the basis of the ATCOR model in a physical way. Tests of the procedure on various imaging spectroscopy data samples show significant improvements of surface reflectance products in forests and in urban areas. Validation results demonstrate the performance and indicate limitations of the proposed methods. Implications for the retrieval of remote sensing products are discussed, with a particular focus on the vegetation indices normalized difference vegetation index (NDVI) and photochemical reflectance index (PRI).
[1]
Rainer Sandau,et al.
Design Principles of the LH Systems ADS40 Airborne Digital Sensor
,
2000
.
[2]
R. Richter,et al.
Geo-atmospheric processing of airborne imaging spectrometry data. Part 2: Atmospheric/topographic correction
,
2002
.
[3]
James A. Gardner,et al.
Algorithm for de-shadowing spectral imagery
,
2002,
SPIE Optics + Photonics.
[4]
G. Asner,et al.
Canopy shadow in IKONOS satellite observations of tropical forests and savannas
,
2003
.
[5]
Paul E. Lewis,et al.
MODTRAN5: a reformulated atmospheric band model with auxiliary species and practical multiple scattering options
,
2004,
SPIE Defense + Commercial Sensing.
[6]
R. Richter,et al.
De‐shadowing of satellite/airborne imagery
,
2005
.
[7]
Rulon Mayer,et al.
Shadowed object detection for hyperspectral imagery
,
2007,
SPIE Optical Engineering + Applications.
[8]
Jens Nieke,et al.
APEX - the Hyperspectral ESA Airborne Prism Experiment
,
2008,
Sensors.
[9]
W. Verhoef,et al.
PROSPECT+SAIL models: A review of use for vegetation characterization
,
2009
.
[10]
Hermann Kaufmann,et al.
Automatic shadow detection in hyperspectral VIS-NIR images
,
2011
.
[11]
Yang Shao,et al.
Shadow detection and building-height estimation using IKONOS data
,
2011
.
[12]
Zhe Zhu,et al.
Object-based cloud and cloud shadow detection in Landsat imagery
,
2012
.
[13]
Deren Li,et al.
Shadow detection for color remotely sensed images based on multi-feature integration
,
2012
.