Dictionary Matrix Model and Calculation Algorithm in Sparse Representation for X-Ray Welding Image Recognition

The characteristics of circular defect, linear defect and noise of sub-arc x-ray images are analyzed here. In order to construct the sparse representation dictionary matrix for x-ray welding image recognition, the dimension of the dictionary matrix is determined by analyzing the correlation curve of x-ray images firstly. A new mathematical model for constructing dictionary matrix is proposed. The new mathematical model is appropriate to sub-arc x-ray images. The mathematical model is solved by using Hopfield neural network, the energy function and solving algorithm are also presented. A dictionary matrix for sparse representation is constructed by using the presented algorithm, and real x-ray images recognition test based on sparse representation shows that the constructed dictionary matrix can decrease the model images and has good robust in recognition.