Selection of multiple natural beacons and satellites using GPU accelerated algorithms.
Type of content
Publisher's DOI/URI
Thesis discipline
Degree name
Publisher
Journal Title
Journal ISSN
Volume Title
Language
Date
Authors
Abstract
Images captured by ground-based telescopes, such as the University of Canterbury Mount John Observatory (UCMJO), are distorted by the effects of atmospheric turbulence. To correct for these effects and improve the resolution of the captured image, a low-order tomographic adaptive optics system is being designed. This system will detect background stars so the effects of atmospheric turbulence can be corrected for, and thereby images of satellites in the foreground restored, by using deconvolution techniques.
To locate the area of the sky in a field of view that contains a background star, a Difference of Gaussians blob detection algorithm is implemented using the OpenCV image processing libraries, the CUDA computing platform, and a general-purpose graphics processing unit (GPU). The algorithm locates the position of both stars and satellites in an image with per-frame execution time reduced between 470% and 1,033% compared to the same algorithm utilising a central processing unit. The increase in speed varies based on factors such as image size and signal-to-noise ratio (SNR).
The implementation has a successful location rate of 100% for images with a high SNR and objects with apparent magnitude lower than 5.85. Objects with apparent magnitude higher than 5.85 are not located in images with a pixel size larger than 1024x1024. Stars with apparent magnitude higher than 5.85 are located when they occupy a higher proportion of the pixels in an image.