Research On The Information Processing Technology On Ultrasonic Scattering Computed Tomography

The ultrasonic technology of computed tomography (CT) has been widely used in thefield of medicine, nondestructive testing of industrial materials and earthquake forecasting,etc.. The echo signal which has been used by the ultrasonic scattering CT consists of severalsignals, the mode not only does not need the back-projection reconstruction, but also canmake up for the deficiency of the linear ultrasonic CT method which, to some extent,attributes to the linear assumption and weak scatteringOn the basis of being familiar with and understanding the result that the research teamhas studied, which acquires signal by virtual phased array and preprocessed signal byspace-time compression. The method of classification of ultrasonic signal has been putforward in this paper, which in a further step can improve the preliminary rapid location of theproperty and size of the defects and the shortage of reconstruction scope. And it makes thefoundation for improving the accuracy and speed of the CT reconstruction. Ultrasonicscattering CT signal is nonlinear and non-stationary, so Hilbert-Huang transform (HHT) hasbeen applied to the signal processing of ultrasonic CT, and then the instantaneous frequency,Hilbert spectrum and marginal spectrum have been obtained. The defects can be classified bythe instantaneous amplitude, the instantaneous phase and the instantaneous frequency thatobtains through the method of HHT, the relative size of defects can be estimated by theinstantaneous amplitude, which can supply a priori information for the signal qualitative andquantitative analysis. Paper adopts the method of conventional analysis; wavelet transformanalysis and Hilbert-Huang transform analysis to extract feature of the pretreatment signal,and takes them as input vector of BP neural network. The purpose of qualitative classification has been achieved through the training and learning on BP neural network. Finally, thepretreatment signal has been quantitative recognition by using method of conventional,wavelet energy-spectrum-coefficient and the Hilbert-Huang transform.By qualitative, quantitative analysis and processing for the defects of CSK-IA titaniumalloy test blocks it has been completed through the method of combining wavelet, HHT andneural network. The result shows that the method which has been put forward in this paper isfeasible and valid for the processing of planar interface.