Optical emission spectroscopy (OES) is a diagnostic for atmospheric-pressure plasmas (APPs), yet its accuracy depends on separating continuum spectra from overlapping line radiation, still done largely by hand. We assessed four automatic algorithms on measured and synthetic APP spectra: two generic signal-processing tools, Hilbert-transform envelope tracing and robust regression with RANSAC, and two plasma-specific routines, tangent-line linkage between successive peaks and recursive interpolation through local minima. The Hilbert method fails to track the slowly varying baseline, and RANSAC helps only when the continuum follows a simple analytic shape. On the other hand, both dedicated routines recover the baseline across the 250-900 nm range and remain robust under realistic noise; the tangent-line scheme yields the most faithful envelope especially when noise levels low enough, while the minima-interpolation variant is faster and highly resistant to noise. Parameter scans with noise-augmented synthetic spectra give practical windows, smoothing levels, orders of local minima and recursion depths. The validated workflow provides consistent continuum-line decomposition, improving subsequent retrieval of electron density and electron energy distribution from OES.
各種検索
サポート
ヘルプ