Consistent Clustering of Radar Reflectivities Using Strong Point Analysis: A Prelude to Storm Tracking

An image segmentation algorithm using an alternating erosion/dilation technique called strong point analysis (SPA) is introduced for general-purpose feature detection. The ability to associate and group pixels with the salient features of an image allows computers to consider images not as an array of values but as a collection of objects. This enables other algorithms to perform advanced tasks, such as tracking an object in a time series of images. The qualitative needs for proper tracking of storm cells in radar images are discussed. To test SPA for those qualities, radar reflectivity images from three S-band weather radars were used. The algorithm is demonstrated to identify features fairly consistently over a time series of images, as well as exhibiting well-behaved changes to its output with respect to changes to the algorithm's input parameters.