Abstract

Multi-temporal images acquired at high spatial and temporal resolution are an important tool for detecting change and analyzing trends, especially in agricultural applications. However, to insure a reliable use of this kind of data, a rigorous radiometric normalization step is required. Normalization can be addressed by performing an atmospheric correction of each image in the time series. The main problem is the difficulty of obtaining an atmospheric characterization at a given acquisition date. In this paper, we investigate whether relative radiometric normalization can substitute for atmospheric correction. We develop an automatic method for relative radiometric normalization based on calculating linear regressions between unnormalized and reference images. Regressions are obtained using the reflectances of automatically selected invariant targets. We compare this method with an atmospheric correction method that uses the 6S model. The performances of both methods are compared using 18 images from of a SPOT 5 time series acquired over Reunion Island. Results obtained for a set of manually selected invariant targets show excellent agreement between the two methods in all spectral bands: values of the coefficient of determination (r²) exceed 0.960, and bias magnitude values are less than 2.65. There is also a strong correlation between normalized NDVI values of sugarcane fields (r² = 0.959). Despite a relative error of 12.66% between values, very comparable NDVI patterns are observed. View Full-Text