Attention structured Pattern Deep Learning for region induction thermography NDT

Induction Thermography is a crucial nondestructive testing technology which has a rapidly increasing range of applications for crack detection. A series of studies have been carried out in thermal sequences processing algorithms. However, these methods often suffer from the complex surface conditions of experiment components. In this study, we propose an end-to-end attention structured pattern deep learning method to achieve precise crack detection and localization. The proposed method integrates both time and spatial pattern mining for crack information with a deep region convolution neural network. Experiments on welding line cracks have shown attractive performance and verified the efficacy of the proposed structure.