New technologies for texture and fractal analysis of multidimensional radar images

Background. A considerable number of fractal features are currently known, each individually characterizing the scaling, singular, directional, and multifractal properties of a texture. Therefore, for solving a range of thematic processing tasks for the same image, it becomes necessary to apply various texture-fractal features, which possess different informative and discriminatory power. From both the theoretical and practical perspectives of digital image processing, this approach fails to meet the efficiency requirements for a broad range of applications. Aim. To substantiate a unified processing method for multidimensional data from aerospace monitoring radar systems. This method is based on fractal features that are universal across different feature spaces of multidimensional and multi-parameter radar images of aerial and ground targets, formed under jamming conditions, with the goal of automating the interpretation process. Methods. The directional morphological multifractal signature method employed in this work is based on the iterative morphological formation of «upper» and «lower» covers using dilation and erosion operations, respectively, with a linear structuring element rotated through a discrete range of angles. The procedure involves calculating the «volume» between the original image and the iterative covers, estimating the surface area, computing the generalized statistical sum as a function of the distribution of the q-th order scaling moment measure of the multifractal set at each analyzed scale, and forming multifractal signatures by determining the generalized statistical sum between adjacent analysis scales, with final value correction based on the predominant texture direction. Results. A concept of multidimensional radar imagery is proposed, with a justification for the maximum possible dimensionality of the data array formed by a synthetic aperture radar. The method allows for the simultaneous estimation of nearly all known fractal parameters (with the exception of lacunarity) while accounting for anisotropy, and generates their corresponding images using a unified scientific and methodological framework. This eliminates information loss associated with the separate computation of all features by different methods and achieves a synergistic effect from the application of this approach. Based on the proposed method, technologies for generating a universal fractal feature for parametric, multiband, and polarimetric radar images have been developed. Conclusion. Despite the substantial computational overhead required to implement the proposed method, this approach enables the use of a universal set of fractal features for the entire diversity of radar imagery. This universality is achieved irrespective of the imagery’s dynamic range and the physical nature of electromagnetic wave scattering across various frequency bands and polarizations, while preserving high reliability and informational content. The results of texture-fractal processing of two-dimensional, multiband, and polarimetric radar images formed by Synthetic Aperture Radar systems demonstrate high reliability and completeness of the extracted information.