Automatic determination of the plasma frequency using image processing on Whisper data

Wave of High frequency and Sounder for Probing of Electron density by Relaxation (WHISPER) performs the measurement of the electron density on the four satellites of the CLUSTER project. The two main purposes of the WHISPER experiment are to record the natural waves and to make a diagnostic of the electron density using the sounding technique. The various working modes and the Fourier transforms calculated on board provide a good frequency resolution obtained in the bandwidth 2-83 kHz and a well instrumental adaptability to determine the electron density in various plasmas. In active mode, spectrograms exhibit various resonance frequencies, the number of which depends strongly of the sounding region. The main purpose of this presentation is to show methods to recognize automatically the plasma frequency resonance Fpe. And, how we identify the different resonance frequencies as upper hybrid frequency Fuh, gyrofrequencies and harmonics Fcen, and Berstein modes Fqn, Interesting resonances which are able to determine the plasma density are not always the most powerful peaks of the spectrum. Sometimes, it is difficult to select the resonance frequencies from erratic peaks. A Spectrogram is considered as a picture and techniques of image processing are developed taking into an account a global information in time and frequency. Directive filters have been adapted to increase the contrast and so define more clearly the resonance frequency pattern along the time. We present results for two different magnetospheric regions: magnetosheath and plasmasphere where the behavior of the active spectrograms is strongly different. Inside the first region, the Frechet distance is used to compare the Whisper data with the EFW potential variations. Whereas in the second region a new picture is completely reconstructed and filtering techniques are used to increase the data quality and identify the most powerful information. Moreover, physical information as dispersion relation, is injected in the data treatment to recognize and name the characteristic frequencies. Lastly, the propagation properties of the natural waves are used to increase the time resolution of the electron density. Examples and statistical validation are showed.