Automated Feature Recognition, Tracking, and Quality Control in Ground-Based Images of the Upper Atmosphere

Optical imaging is a valuable means of observing upper atmospheric structures, but tropospheric cloud cover and lack of automated means of data analysis present significant barriers to operational use of data from ground-based optical instruments. We have developed several concepts for identification and tracking of equatorial plasma depletions in all-sky imager data under realistic conditions including significant cloud cover. The most successful algorithm relies on identification of depletions by their motion, and allows multiple frames of image data to be processed coherently along a preferred axis in a 3-D spatio-temporal data structure corresponding to the depletion motion. This greatly reduces cloud effects up to 50% cloud cover. When applied to the background star field, our correlation-based technique collapses a random star pattern into a single easily identifiable peak in a correlation distribution. The displacement of this star peak in a cross correlation provides a robust means to automatically distinguish between true stars and hot pixels or noise, and allows rapid automatic identification of image regions affected by clouds without the need to identify, locate, or track any specific stars.