Deep learning and Bayes data fusion for crack detection

摘要：

　　Regular inspection for the components inside nuclear power plants is important to ensure their safety.However,current inspection practice requires operators to watch video for identifying cracks on the component surfaces,which is time consuming,tedious,and subjective.Moreover,a few autonomous crack detection approaches for nuclear power plant inspection have been developed.The existence of scratches,welds,and grind marks on the component surfaces leads to a large number of false positives when prevalent crack detection approaches are used.This study proposes a new framework to detect cracks in videos that consists of a deep convolutional neural network(CNN),a spatiotemporal registration procedure,and a Na?ve Bayes data fusion scheme.The crack patches in different frames are registered together based on their spatiotemporal coherence,and the posterior probabilities of being real cracks are derived.The proposed approach achieves 98.3%hit rate with only 0.1 false positive-per-frame which is superior to the previous studies.

关键词： crack detection convolutional neural network data fusion nondestructive testing structural health monitoring

作者: Fu-Chen Chen Mohammad Reza Jahanshahi

作者单位: Electrical and Computer Engineering,Purdue University,West Lafayette,USA Civil Engineering,Purdue University,West Lafayette,USA

会议类型: 国际会议

会议名称: The 7th World Conference on Structural Control and Monitoring(7WCSCM)(第七届结构控制与监测世界大会)

会议地点: 青岛

会议语种:英文

页码: 1536-1543

在线出版日期: 2018-07-22（万方平台首次上网日期，不代表论文的发表时间）