SAR Image Simulation and Similarity Evaluation based on Basic Structure of Buildings

Abstract

SAR images plays a critical role in observing and understanding buildings in urban areas. Yet traditional image databases come with a high cost. It is therefore highly desirable to develop a novel approach to achieve realistic simulated SAR images. In this paper, a simulation method focused on the geometrical features of buildings is introduced. The intensity of the simulated image is obtained by an improved Phong-Shading model. And a ray tracer software, Pov-Ray, is used to establish the mapping between the 3D scene and the 2D SAR images. Then, post processing is introduced to render the SAR images more realistic. The proposed method is evaluated by assessing the similarity between real images and simulated images. Experimental results demonstrate that the proposed method yields a high similarity in the geometrical structures. Various geometrical targets can be combined together to make a complex object. This modular approach can establish a database with a great number of simulated SAR images.