Radiometric Terrain Correction of SPOT5 Image

Xiuli Feng,Feng Zhang,Ke Wang

DOI: https://doi.org/10.1117/12.760109

2007-01-01

Abstract:Remote sensing SPOT5 images have been widely applied to the surveying of agriculture and forest resources and to the monitoring of ecology environment of mountain areas. However, the accuracy of land-cover classification of mountain areas is often influenced by the topographical shadow effect. Radiometric terrain correction is important for this kind of application. In this study, a radiometric terrain correction model which based on the rationale of moment matching was made in ERDAS IMAGINE by using the Spatial Modeler Language. Lanxi city in China as the study area, a SPOT5 multispectral image with the spatial resolution of 10 m of that mountain area was corrected by the model. Furthermore, in order to present the advantage of this new model in radiometric terrain correction of remote sensing SPOT5 image, the traditional C correction approach was also applied to the same area to see its difference with the result of the radiometric terrain correction model. The results show that the C correction approach keeps the overall statistical characteristics of spectral bands. The mean and the standard deviation value of the corrected image are the same as original ones. However, the standard deviation value became smaller by using the radiometric terrain correction model and the mean value changed accordingly. The reason of these changes is that before the correction, the histogram of the original image is represented as the 'plus-skewness distribution' due to the relief-caused shade effect, after the correction of the model, the histogram of the image is represented as the normal distribution and the shade effect of the relief has been removed. But as for the result of the traditional C approach, the skewness of the histogram remains the same after the correction. Besides, some portions of the mountain area have been over-corrected. So in my study area, the C correction approach can't remove the shade effect of the relief ideally. The results show that the radiometric terrain correction model based on the rationale of moment matching is an effective model to reduce the shade effect than the traditional C correction approach, especially in the complex undulation of mountain area with lots of shade effect. In other words, the traditional C correction approach will show the better result at the plain area with less shade effect. Besides, the accuracy of the DEM data and the registration accuracy between the image and the DEM data will also influence the final correction accuracy. In order to achieve the higher radiometric terrain correction, high spatial resolution DEM data is preferred.