Position-independent Classification For Multiple Stationary Targets Embedded In Closed Environment

With the development of the technology of through-the-wall radar detecting, UWB radar has been widely used in military and civilian, because of its high range resolution. Ultra-wideband through-the-wall radar imaging is a process of intrusion detection and imaging the targets hidden in the wall. Usually the targets after the wall are not identical in size, shape character or dielectric constant. And signal bandwidth is limited due to the attenuation of the wall, which cause the difficult of the understanding and recognition of the characteristics of targets. Even if the background medium and the dielectric constant of the targets are all the same, the statistical features of images also varies because of the target location. Aiming at this problem, this paper based on the narrow pulse ultra-wideband through-wall radar as a means of detection, from the perspective of Green’s function, realized position-independent target classification in the indoor complex scenario.At first, Back Projection imaging is used to realize the targets imaging of different position. Through the analysis of the propagation path of electromagnetic wave and the rapid wall compensation algorithm, realize the target fast single perspective imaging. After imaging, radar image and the targets classification performance are related to the imaging system(such as the work frequency of the antenna), size, shape and performance parameters of targets, and the position of the targets and antenna. Then, combined with green function two-dimensional deconvolution target contrast function spatial distribution analytical expression is deduced, the targets aligned to the reference position. Thus decoupling target characteristics affected by target and antenna relative position information. When the frequency of the antenna and the background parameter are the same, multiple factors are constrained to the target parameter only. At last, extract the feature reflecting targets’ electrical characteristics from the aligned targets. Use Support Vector Machine(SVM) to achieve robust indoor static multi-objective classification.This article innovative based on BP and linear Born Approximation electromagnetic inverse scattering mathematical model deduce the mathematical model of systematically achieving robust indoor static multi-objective classification with different positions.