A Detection Method Of Infrared Small Target Based On EMD

Hilbert-Huang transformation is a kind of recent signal analysis theory which is first proposed by NASA Norden E. Huang et al at the end of last century. It is based on local characteristics of the signal. It can carry on auto-adapted and highly effective decomposition to the signal. Moreover it is suitable especially for the analysis of non-linear and non-stationary. It has important theory value and broad application prospect.In this paper, the Hilbert-Huang Transform (HHT) has been introduced as a new method. Starting from the concept of the instantaneous frequency, the intrinsic mode functions which made the instantaneous frequency meaningful, and the way how to realize empirical mode decomposition (EMD) has been introduced. The final presentation of HHT result is in 3D of time-frequency-energy/amplitude distribution, which is defined as the Hilbert spectrum.Empirical mode decomposition is an essential part of Hilbert-Huang transformation. In the paper, four important issues of the empirical mode decomposition are discussed. Four important questions of how to process the empirical mode decomposition if one-dimensional EMD is promoted to bidimensional EMD. The method of promoting the one-dimensional EMD to the bidimensional method based on the Delaunay triangulation and cubic polynomial interpolation discussed emphatically.The detection, recognition and tracing of infrared small target under complex backgrounds is a very important and challenging task in the application of modern military affairs. The paper proposes a method in a creative way that DEMD which is based on the Delaunay triangulation and cubic polynomial interpolation is applied in extracting the infrared small target under complex background. At the same time, a comparative research on DEMD, BEMD which is based on the tensor indicates that DEMD is better than BEMD in the aspects of examinational capability and the iterative time it costs. Through the comparison of DEMD, BEMD and the wavelet transformation, it confirms that EMD has advantages over wavelet transformation in the aspects of dealing with non-linear and non-stationary signals.