A dynamic star spots extraction method based on pixel association

Abstract

Star trackers are devices that determine high-accuracy attitude by observing stars. As aerospace technology develops, the dynamic performance of the star tracker becomes increasingly important. In the star images taken under high dynamic conditions, the star spots will be tailed or fractured. It will cause a sharp decrease in Signal-to-Noise Ratio (SNR) and make it a challenge to extract and position star spots. In this paper, an Object-to-Pixel star spots extraction and positioning method based on pixel association was proposed, which is a breakthrough compared to the existing Pixel-to-Object extraction method. In the proposed method, rough extraction of star spots is carried out by the Radon transform and image entropy. Precise positioning of star spots is carried out by Gaussian mixture model (GMM) and maximum likelihood estimation (MLE). Experimental results indicate that the proposed method can achieve reliable extraction for star spots with SNR less than 3 or even close to 1. Meanwhile, the positioning accuracy can be improved by at least twice compared with the conventional threshold method.