Main content area

A Spatial-Spectral Methodology to Detect Narrow Shadows on Satellite Imagery

Zhang, Ying, Guindon, Bert
Photogrammetric engineering and remote sensing 2013 v.79 no.3 pp. 269-276
buildings, cities, image analysis, radiometry, remote sensing, suburban areas, Alberta
<p><i>Image shadow is an important mono-scope cue in the detection and characterization of buildings from aerial and satellite imagery. Previous studies have confirmed the effectiveness of radiometric thresholding in delineating shadows cast by tall buildings on high resolution (≤1m) imagery. Thresholding is not effective when shadow width is comparable to the imaging sensor resolution since mixed shadow/illuminated pixels do not exhibit a distinct radiometric character. A spatial-spectral shadow detection method is proposed that exploits the line-like and low radiance characteristics of shadows cast by large area commercial/industrial buildings commonly found in suburban areas of large cities. Tests conducted in shadow detection of such buildings in Calgary, Alberta using RapidEye imagery indicate that detectability rates of ~80 percent can be achieved although commission errors are significant because of the plethora of other line-like features present in industrial park settings.</i></p>