AN ADAPTIVE CONSTANT FALSE ALARM RATE DETECTION SYSTEM

摘要

A method and apparatus of Constant False Alarm Rate (CFAR) detection resultsin optimaldetection in the Neyman Pearson sense independent of the type of the inputdata. The whole systemconsists three major blocks: pre-processing block, data fusion block andadaptive multi-CFARdetection block. For any kind of input data, various appropriate dataprocessing techniques could bechosen and applied in parallel in the pro-processing unit for the purpose ofenhancing Signal toNoise Ratio (SNR) and fusion. The SNRs of the output data from the pre-processing block will becompared and the top couples will be transmitted to the data fusion block forfurther SNRenhancement by performing appropriate data fusion technique. The adaptivemulti-CFAR detectorwill then be applied at the last stage to get the final decision.In the application of this system on hyperspectral data, a new Transform BasedMatched Filter(TBMF) detection algorithm is proposed in the pre-processing unit. The TBMFalgorithm employs atransformation before the traditional matched filter to make the backgroundsubspace and targetsubspace orthogonal, hence enhance the detection performance of the matchedfilter algorithm.