Identification of clouds and aurorae in optical data images

In this paper we present an automatic image recognition technique used to identify clouds and aurorae in digital images, taken with a CCD all-sky imager. The image recognition algorithm uses image segmentation to generate a binary block object image. Object analysis is then performed on the binary block image, the results of which are used to assess whether clouds, aurorae and stars are present in the original image. The need for such an algorithm arises because the optical study of particle precipitation into the Earth's atmosphere by the Ionosphere and Radio Propagation Group at Lancaster generates vast data-sets, over 25 000 images/year, making manual classification of all the images impractical.