Detection of fallow lands overgrowth from satellite images based on wavelet analysis

Abstract

To improve the efficiency of agricultural lands condition and use monitoring, method for detecting fallow lands overgrowth with tree and shrub vegetation using multi-temporal multispectral satellite images of high spatial resolution is proposed. The task of localization of overgrowth is solved on the basis of multi-temporal satellite images wavelet analysis. This technique allows to operationally and automatically detect areas of fallow lands overgrowth with tree and shrub vegetation. The proposed methodology was tested using satellite images obtained from Quickbird, GeoEye and WorldView satellites. For each image, a fourth-order wavelet decomposition was performed in terms of the Daubeshi wavelet functions. Comparing the different levels of the images wavelet transform, it is possible to divide the areas without tree and shrub vegetation, and where overgrowth was depicted. The method proposed by the authors for detecting changes based on wavelet analysis has shown that this method practically does not depend on the influence of shooting conditions and, accordingly, is more effective at detecting changes.