The Research Of Underwater Target Scattering Feature Based On Time-Frequency Analysis

The technology of the underwater target detection and recognition has been one of the field that many scholars focused on. When the active sonar detected underwater target, the analysis of echoes is the best effective method. The echo is generated by the interaction between the incidence sound wave and the target, including a lot of the physical information of the target.This is an important source to detect and identify underwater targets. According to the target echoes highlight model theory, the characteristic of underwater target echo signal is affected by the incident angle. In order to conduct intensive study on the target, it is not only need to extract the feature of the target echoes, but also to establish the feature library of the echoes under all-direction incident angles. Aiming at the above problems, in this paper, a variety of time frequency analysis method is used to extract the feature of the target echo. This paper establishes the feature library of the echoes under all-direction incident angles and does a further research on the target recognition.Firstly, this paper introduces the scattering characteristics of the underwater target and the mechanism of the target echo. And the highlights characteristics of the target echo relationship with the sound wave incident angle is studied, and then simulates the relationship. The time-frequency characteristic of the target of echoes signal is studied, and then the Wigner-Ville transform and fractional Fourier transform of the target geometric highlights echoes signal are derivation. The result of processing the experimental data by WVD-Hough Transformation, FRFT and frequency domain WT demonstrates that these three methods can extract the energy characteristics, highlight statistical characteristics and elastic highlight characteristics of target echo effectively, and the characteristics of target will change with the incident angle of sound waves. Then establish the feature library of omnidirectional target echo in respective transform domain by WVD-Hough Transformation, FRFT and Frequency Domain WT. Dimensionality of the primitive feature vectors in respective space are usually very high, it's hard to classify and recognize the target directly, making a dimension deducing is necessary. The highlight characteristics of underwater target echo is researched in this paper,which are the numbers of highlight, locations between each highlight, and intensity of highlight. Then the time-frequency features of target are choosen and applied the feature vector of classifier. Each character of target echo is the discription in respective feature space,and there's complementarity between different feature spaces. A modified feature level fusion is adopted in this paper to fuse the processed feature vector, then recognize and research the gesture of target by the fused features. The precessing result of the obtained omnidirectional target data demonstrates that it is effective to establish the characteristic model of omnidirectional target echo and choose the several characteristics of target echo highlight in target recognition. Furthermore, the fused features raise the recognition accuracy of target.