Search

Remote sensing radar techniques provide highly detailed imaging. Nevertheless, radar images do not offer directly retrievable representations of shape within the scene. Therefore, shape reconstruction from radar typically relies on applying post-processing computer vision techniques, originally designed for optical images, to radar imaging products. Shape reconstruction directly from raw data would be desirable in many applications, e.g. in computer vision and robotics. In this perspective, inversion seems an attractive approach. Nevertheless, inversion has seldom been attempted in the radar context, as high frequency signals lead to energy functionals dominated by tightly packed narrow local minima. In this paper, we take the first step in developing a framework in which radar signals and images can be jointly used for shape reconstruction. In particular, we investigate the feasibility of shape reconstruction by inversion of pulse-compressed radar signals alone, collected at sparse locations. Motivated by geometric methods that have matured within the fields of image processing and computer vision, we pose the problem in a variational context obtaining a partial differential equation for the evolution of an initial shape towards the shape-reflectivity combination that best reproduces the data. While doing so, we highlight several non-obvious difficulties encountered and discuss how to surpass them. We illustrate the potential of this approach through three simulated examples and discuss several implementation choices, including boundary conditions, reflectivity estimation, and radiative models. The success of our simulations shows that this variational approach can naturally accommodate radar inversion and has the potential for further expansion towards active surfaces and level set applications, where we believe it will naturally complement current applications with optical images. [ABSTRACT FROM AUTHOR]

Image-driven data mining methods are described for image content segmentation, classification, and attribution, where each pixel location of an image-under-analysis is the center point of a pixel-block query that returns an estimated class label. Feature attribute estimates may also be mined when sufficient attribute strata exist in the data warehouse. Novel methods are presented for pixel-block mining, pattern similarity scoring, class label assignments, and attribute mining. These methods are based on a direct sum tree structure called a σ-tree that is utilized with near-neighbor similarity scoring. The σ-tree structure provides a solution to the challenge of high computation/memory costs of pixel-block similarity searching. The σ-trees are integrated into warehouse subsystems that provide referential capability into feature attribute data, resulting in a foundation for data mining called Source Optimized, Labeled, DIgital Expanded Representations (SOLDIER). The variable depth "bit-plane" data representations produced by σ-tree path selections provide an approach to image content segmentation, and provide a structure for formulation of Bayesian classification with data-adaptive Parzen classifiers with variably sized windows. Preliminary methods and results for postprocessing of mined feature-thematic layers for higher level scene understanding are also presented. Sample results are shown with synthetic aperture radar images and with high-resolution pan-sharpened satellite images of the Payagala, Sri Lanka area before the site was devastated by the 2004 Asian Tsunami. [ABSTRACT FROM AUTHOR]