Classification of Impulse Radar Data Using Neural Networks with Experimental Data(Klassificering av Impulsradardata med Neurala Naetverk med Experimentella Data)

摘要

The report describes a neural network based approach for classification ofobjects in air and soil. The report presents how tools for neural networks have been applied to data achieved from measurements of radar pulses towards objects in the air (model aircrafts) and in ground (buried mines). In each environment, there were a small number of object types (models and mines respectively). The methodology is based on a two step arrangement where phase one is net design (learning step) and phase two classification. In each phase, the frequency specta of typical objects were used. The authors found that it is possible to automatically build neural nets that learn to recognize the shapes of typical objects in air and ground respectively, an ability that later was used for classification of such objects.