Abstract

Several machine learning (ML) classifiers were tested to automate the analysis of time-domain terahertz spectroscopy (THz-TDS) imaging, which is inherently the hyperspectral technique with often large datasets having the sliced image data in frequency and time domains. Among the tested ML classifiers for THz waveform recognitions, the random forest one was the best choice in terms of overall accuracy, speed, robustness, and user-friendliness (see Table 1 and Fig. 1). Combining these and our previously reported results in frequency domain [1], such ML techniques could have a great potential for THz-TDS analytical and non-destructive testing (NDT) applications.

© 2019 Japan Society of Applied Physics, The Optical Society (OSA)

PDF Article

References

You do not have subscription access to this journal. Citation lists with outbound citation links are available to subscribers only. You may subscribe either as an OSA member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access OSA Member Subscription