HF band emitter detection and segmentation based on image processing

An algorithm for automatic segmentation of wide band spectra is presented. The algorithm is based on the interpretation of the angle spectrum as a set of binary images. Therefore, only the time frequency spectrum of angles and not the power spectrum is used for the emitter detection. This leads to advantages in the separation of different emitters which are overlapped in the time-frequency domain, as well as of spread spectrum emitters. The segmentation is implemented with image processing algorithms applying classical erosion and dilatation procedures with various structure elements. In detail the different processing steps are described to obtain a symbolic description of the emitters, e.g. the time-frequency description or the azimuth angle. These parameters can be used to get a rough classification of different emitter types. Some examples are shown with real data in a simulation environment of a wide band location system.