Non-stationary Signals Generalized S Transform And Its Application In The Sar Image Analysis Study

As a useful tool to analysis non-stationary signal, time-frequency analysis is one of the hot issues in modern signal processing field. Stockwell transform (ST) is a new method proposed and has lots of application in 1D signal processing and little application in 2D signal processing so far. Because SAR signal is a complex non-stationary signal, a suitable time-frequency analysis method is the key technique for SAR image processing and analysis. At present, the research on image processing and analysis method to SAR is still in its starting stage.This paper first puts forward a method which takes generalized Stockwell transform (GST) as a new and effective tool for time-frequency spectral analysis of SAR image signal. The research in this aspect has significant value in the theoretical meaning and practical application.This paper is started from the basic linear time-frequency theory. As follow, the time-frequency analysis methods of short-time Fourier transform, wavelet transform and ST are introduced briefly. Moreover, comparison of advantages and disadvantages and relationship are made among them.The basic principle and algorithm implementation of ST are the keystone of this paper. In order to improve practicability and adaptability of ST, GST which has adjustable time-frequency resolution is obtained by adding two adjustable parameters to improve the Gaussian window function. Meanwhile, the application principle and algorithm implementation of GST in image processing are also stated in detail.Finally, simulation experiments are obtained combined with a simulation model and two image data of actual scene. The simulation model of Linear Frequency Modulation Signal (LFM) proves the effectiveness and feasibility of ST in the non-stationary signal processing. The comparative experiments of SAR images and optical images based on the same scene demonstrates that ST is more suitable for SAR image processing which has abundant time-frequency information compared with optical images; The comparative experiment of ST and GST which has an adjustable time-frequency resolution based on the same SAR image demonstrates that GST spectrum can provide more complementary time-frequency information than ST spectrum and it can describe the localization time-frequency feature of SAR image signal more exactly.