The charge-coupled device camera of the TJ-II Thomson scattering (TS) can capture five different classes of images. Typically, different data processing is performed depending on the kind of image that is acquired. The procedure can be automated to recognize the type of image. To this end, machine learning methods (MLM) are applied. However, usually, MLM classify without confidence estimates. An image classifier based on conformal predictors has been developed for the TJ-II TS. It provides a couple of indicators (confidence and credibility) for each classification that measures the accuracy and reliability of the prediction. Results achieve success rates of about 97%. The implemented classifier is valid for any kind of images. © 2010 American Institute of Physics.
All Science Journal Classification (ASJC) codes
Vega, J., Murari, A., Pereira, A., González, S., & Pastor, I. (2010). Accurate and reliable image classification by using conformal predictors in the TJ-II Thomson scattering. Review of Scientific Instruments, 81(10), -. [10E118]. https://doi.org/10.1063/1.3478689