Multifeature Fusion Imaging Based on MachineLearning for Weld Defect Detection UsingInduction Scanning Thermography

The detection of weld structures in service issignificantly hindered by surface irregularities and blem-ishes, presenting a challenge to detection technology.To address this issue, the present study proposes a detectionplatform that integrates magnetic yoke excitation with induc-tion scanning thermography (IST) technology for heatinglarge-scale and complex structures. After the collection andpreprocessing of thermography sequence data, a multifea-ture fusion imaging method is devised based on multipletime-domain thermal feature images input and reevaluatesthe effectiveness of these individual features in distinguish-ing defects from backgrounds. Furthermore, an improvedphase spatial kurtosis (PSK) feature that utilizes dynamicdetection signal mechanisms is proposed. The aforementioned features are combined through a trained artificial neuralnetwork (ANN) in order to obtain an estimation. The experimental results demonstrate that the proposed methodeffectively detects and isolates the defect area in complex backgrounds at the pixel level, enhancing imaging accuracyto 99.6% while significantly suppressing interference from complex background information without generating defectfalse positives (FPs)