Calibrated ratio approach for vegetation detection in shaded areas

Abstract

Abstract. Removing shadow effects remains a challenge in processing optical remote sensing data. Shadows occur because of obstructions from the terrain topography or cloud cover, which can cause errors for image classification. Shadow effects can be removed using a band-ratio approach because the shaded areas in optical images have a nearly proportional variation in the bands. We developed a calibrated band-ratio approach for shadow reduction. Before the ratio approach was applied, a regression technique was used to obtain information for calibration from the relative sensor gain and offset. After calibration, the ratio vegetation index (RVI) band ratios were calculated to process the image data, which can simultaneously remove the shadow effects and assist the search for vegetation. Real and synthesized images show that the calibratedratio approach can improve vegetation detection compared with standard RVI and dark pixel subtraction approaches.

Unsupervised deep learning of landscape typologies from remote sensing images and other continuous spatial data

“…We chose to take the ratio between colour bands (i.e. (band1 -band2)/(band1 + band2)), because, compared to raw images, these tend to be less biased by shadow effects of mountains (Bijeesh and Narasimhamurthy, 2020;Kao et al, 2014), which are prevalent in Switzerland. Such ratios are also recommended to "eliminate illumination changes" (Ball et al, 2017, p. 19), which may be present in mosaics composed of satellite images taken on different dates and times.…”

Illumination invariance and shadow compensation via spectro-polarimetry technique

“…The energy is captured in a narrow slice of wavebands of approximately [5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20] nm. Each recorded pixel in each waveband contains the spatial and spectral information that can be extracted as target reflectance or signature as a function of wavelength.…”

“…A large amount of research has also been reported for illumination invariance and shadow compensation on hyperspectral imageries. Band ratio and normalization techniques were the first approaches introduced (16,17). These techniques are simple but strongly dependent on the variation of targets in the image.…”