A novel feature extraction and enhancement technique for infrared thermal wave imaging based on multi-time local outlier factor

• Published in: Journal of thermal analysis and calorimetry Vol. 150; no. 16; pp. 12401 - 12413
• Main Authors: Yin, Peng, Wang, Fei, Zhou, Yulong, Du, Qihou, Li, Rongcheng, Yue, Honghao, Liu, Junyan
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.08.2025; Springer Nature B.V
• Subjects: Air flow; Analytical Chemistry; Chemistry; Chemistry and Materials Science; Comparative analysis; Composite materials; Defects; Excitation; Heat sources; Information retrieval; Infrared imaging; Inorganic Chemistry; Measurement Science and Instrumentation; Physical Chemistry; Polymer Sciences; Signal to noise ratio; Thermal imaging; Thermal wave imaging; Data processing algorithm; Feature extraction algorithm; Infrared thermal wave imaging; Local outlier factor
• ISSN: 1388-6150; 1588-2926
• DOI: 10.1007/s10973-025-14446-8

This paper proposes an infrared defect information extraction and enhancement algorithm based on multi-time local outlier factors(MT-LOF), which is designed to address the problem of background information interference caused by thermal excitation dominated by Gaussian heat sources. Initially, a series of composite material specimens were fabricated and flaws were emulated through the use of flat-bottomed apertures. Secondly, a nonuniform thermal excitation detection device was constructed, which simulates nonuniform thermal excitation using hot air flow. Finally, experimental research was conducted on defect signal extraction and defect information enhancement. The comparative analysis of locked in thermal imaging (LIT) processing results shows that MT-LOF demonstrates remarkable anti-interference capabilities. In terms of signal enhancement, a comparative analysis of the processing outcomes derived from three specimens reveals that the signal-to-noise ratio of the MT-LOF approach exhibits an enhancement of over 100% in comparison with the LIT method. Additionally, the background signal fluctuation demonstrates a substantial improvement of 70.46%.