Multi-perspective imaging and image interpretation

High resolution range profiling and imaging have been the principal methods by which more and more detailed target information can be collected by radar systems. The level of detail that can be established may then be used to attempt classification. However, this has typically been achieved using monostatic radar viewing targets from a single perspective. In this chapter methods for achieving very high resolutions will be reviewed. The techniques will include wide instantaneous bandwidths, stepped frequency and aperture synthesis. Examples showing the angular dependency of high range resolution profiles and two-dimensional imagery of real, full scale targets are presented. This data is examined as a basis for target classification and highlights how features observed relate to the structures that compose the target. A number of classification techniques will be introduced including statistical, feature vector and neural based approaches. These will be combined into a new method of classification that exploits multiple perspectives. Results will be presented, again based upon analysis of real target signatures and are used to examine the selection of perspectives to improve the overall classification performance.