A simulation-based approach towards automatic target recognition of high resolution space-borne radar signatures

Specific imaging effects that are caused mainly by the range measurement principle of a radar device, its much lower frequency range as compared to the optical spectrum, the slanted imaging geometry and certainly the limited spatial resolution complicates the interpretation of radar signatures very strongly. Especially the coherent image formation which causes unwanted speckle noise aggravates the problem of visually recognizing target objects. Fully automatic approaches with acceptable false alarm rates are therefore an even harder challenge. At the Microwaves and Radar Institute of the German Aerospace Center (DLR) the development of methods to implement a robust overall processing workflow for automatic target recognition (ATR) out of high resolution synthetic aperture radar (SAR) image data is under progress. The heart of the general approach is to use time series exploitation for the former detection step and simulation based signature matching for the subsequent recognition. This paper will show the overall ATR chain as a proof of concept and will go into more detail for the part of simulation based recognition. Focusing on space borne sensors, the all-day and -weather capabilities of radar systems coupled with their high resolution capacity through the SAR principle, provides an ideal instrument for remote sensing. Especially the performance of change detection by using repeat-pass image acquisitions over long periods of time is outstanding if compared to optical sensors. The first important step for ATR is to detect potential targets from the image data, which already is a complicated task when using only single image acquisitions. Automatically detecting candidates for targets from multiple image acquisitions can be done in a much more robust way. Additional to the illustration of this detection approach, thoughts on the incorporation of other a priori information sources e.g. digital surface models, maps and optical reference data will be given. After extracting potential target signatures the final decision of classification is performed by a simulation based recognition method. After selecting available 3D model data that come into question for the extracted candidate, signatures are estimated with very efficient self-developed simulation techniques. Afterwards, prominent scattering centers are extracted from the simulated and real signatures by sophisticated filter methods. Feature-based and correlation-based methods will be compared in the final recognition step. The proof of concept will be illustrated on data takes of an airport scene, through operating the TerraSAR-X satellite in its staring spotlight imaging mode, giving high resolution signatures with a range resolution around half a meter and a cross-range resolution around a quarter of a meter. It will be shown that for large-scale aircrafts, a promising performance in recognition can be achieved without using usual machine learning approaches