研究目的
To enhance quantitative defect information in non-destructive testing using a new feature extraction method based on thermography.
研究成果
The proposed local sparse and low rank feature extraction method based on image entropy and fusion enhances defect features in ECPT, reserving more defect information and suppressing background influence better than ICA and RPCA.
研究不足
The selection of range of frames from initial sequence influences the results due to different thermal distributions and contrasts.
1：Experimental Design and Method Selection:
The methodology involves entropy-based image selection, local sparse and low rank decomposition (LSLD), and image fusion to enhance defect features.
2：Sample Selection and Data Sources:
Two experimental samples, an aluminum alloy block with groove defect and a stainless steel sample with a slot defect, were used.
3：List of Experimental Equipment and Materials:
Induction heater, rectangular coil, IR camera.
4：Experimental Procedures and Operational Workflow:
The IR camera records the surface thermal distribution of the sample. Selected images are processed using entropy gradient, RPCA for noise reduction, and LSLD for feature extraction.
5：Data Analysis Methods:
Image fusion is used to integrate features, and metrics like background suppressing degree (BSD) and defect enhancing degree (DED) are calculated.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
