Maximum likelihood wavelet fusion for aerospace NDE applications

Wavelet-based data fusion techniques have been used extensively in several application areas, including quantification of flaw parameters in nondestructive evaluation applications. The use of hybrid systems including the use of wavelets and neural networks for fusion and corrosion characterization is relatively recent. While the hybrid approach usually results in fairly accurate results, the selection of the fusion parameters is manually accomplished and is typically a time-consuming process. This paper proposes a multiresolution data fusion algorithm to improve the performance of corrosion quantification in aircraft lap joints. A maximum likelihood estimation scheme is proposed to fuse data in the wavelet domain. The proposed algorithm reduces the dimensionality of the input and improves the robustness of the network. Initial results indicate that the proposed approach gives similar results to those obtained using manual parameter selection, and indicate the feasibility of the algorithm.