Image understanding of oblique ionograms: the autoscaling problem

The paper presents an image understanding system for oblique ionograms, which are a two dimensional measurement of the high frequency properties of the ionosphere. The system for image understanding of oblique ionograms (called “autoscaling”) presented here involves three stages. Firstly, the ionogram is filtered by estimating the noise in each pixel of the image by fitting orthogonal second order multinomials to its 3 by 3 neighbourhood. Secondly, feature vectors for each trace of the filtered ionogram are extracted using a process of thinning, trimming, model fitting, and a final merging stage that accounts for broken traces. The model used for ionogram traces is an implicit quartic equation, and is fitted to the cleaned up skeleton of the trace using orthogonal distance regression. A third stage of the overall system uses the feature vectors to identify the prominent modes and other important features of the ionograms, and to track these modes and features as they evolve across sequences of ionograms. This work will automate the task of identifying and tracking ionospheric propagation modes in oblique ionograms. The resulting system will be used to process current data streams to extract large volumes of significant information: this will help the understanding of ionospheric processes and can be used as input to automatic frequency management systems for communications networks