A segmentation method of Ultrasonic CT image based on Wavelet Neural Network

Abstract

Due to the characteristic of the intuitivism of image, Ultrasonic tomography technology in non-destructive testing of concrete structure is attached importance. The combination of Wavelet and Neural Networks will constructs a kind of wavelet neural network, which can be used to study the segmentation method of Ultrasonic CT color image. Instead of the traditional Sigmoid function, Wavelet basis function can organically fuse the good time-frequency domain feature and the adaptive advantage of neural network, and overcome the limitations of the slow convergence rate and easily felling into the local minimum of BP neural network.