Automated identification of high intensity regions of interest in spectrograms for analysis of plasma wave measurements

Regions of high intensity in plasma fluctuation spectrograms often indicate wave activity in the plasma. A framework is developed based on the max-tree image filtering algorithm for automatic identification of these high intensity regions of interest (HIROI). Potential applications include automated signal denoising and decomposition, as well as analysis of fluctuation data at scale for validation of wave theory or exploration of wave parametric dependencies. The max-tree algorithm is used to hierarchically organize all possible HIROI in a spectrogram in a tree data structure based on natural nesting relationships, where HIROI with lower threshold contain HIROI with higher threshold. Tree-walking algorithms quickly and efficiently calculate statistical properties for every HIROI, such as extent and total intensity, that are used in deciding which best exhibit the expected characteristics of a wave. A method is developed using a gain function to automatically select a set of non-overlapping HIROI. All possible HIROI are ranked according to the gain function and lower rank HIROI are discarded if they contain, or are contained within, higher rank HIROI. Different gain functions are developed and compared with synthetic data, as well experimental measurements.