A data fusion framework for multiple nondestructive inspection images

Multiple nondestructive inspection (NDI) techniques were applied to detect and quantify the hidden corrosion in aircraft lap joints. The inspection data is presented in raster-scanned images. In this chapter, the use of Dempster-Shafer (DS) theory to fuse multi-frequency and pulsed eddy current inspection data is presented. The NDI images are first discriminated by iteratively trained classifiers. The basic probability assignment (BPA) is defined based on the conditional probability of information classes and data classes, which are obtained from a teardown inspection followed by a X-ray thickness mapping process and the NDI measurements respectively. The DS rule of combination is applied to fuse multiple NDI inputs. The remaining thickness is estimated by applying locally weighted regression of the DS-fused results.