Star map wavelet denoising method based on local anomaly factor

Abstract

The invention provides a star map wavelet denoising method based on a local abnormal factor, which combines wavelet transform with a local abnormal factor LOF algorithm, and applies an abnormal valuedetection algorithm to wavelet denoising on the basis of wavelet transform, thereby realizing star map denoising and star point extraction.

Description

technical field [0001] The invention relates to the technical field of star map recognition, in particular to the technology of star map denoising. Background technique [0002] Due to its high measurement accuracy, InGaAs star sensors are widely used in near-Earth satellites, deep space exploration, ballistic missiles and small satellites and other all-weather applications. For these applications, fast and reliable all-time star sensor star pattern recognition algorithms are critical. The stray light such as the atmosphere and sunlight in the starry environment will directly affect the size of the light spot formed by the star on the star map and the energy distribution of the star point. Among them, the stray light may enter the star sensor after multiple reflections and form on the star map. The light spot changes the position and structure of the stars in the field of view and affects the star point positioning. Therefore, in the star map recognition, the denoising of ...

AI Technical Summary

Problems solved by technology

[0004] 1) The traditional image processing method mainly removes image noise through mean filtering, median filtering, Gaussian low-pass filtering or Wiener filtering, but the star targets in the star map are distributed in a point shape, which has a high similarity with the noise distribution. These traditional image denoising methods also lose the original information of stars while removing noise, which has obvious limitations

At present, there are also some algorithms that have improved various traditional filtering methods, such as a mean filtering method based on over-complete sparse representation proposed by Beihang Mingyuan Zhou et al., which improves the reliability of star point extraction; Wang Min et al. An energy function-based extreme value median filter denoising algorithm is proposed, which can better deal with salt and pepper noise; Nannuo et al. proposed a star image denoising method based on spatio-temporal correlation, through continuous multi-frame Positioning effectively solves the positioning of star maps with low signal-to-noise ratio; and a non-local mean denoising algorithm based on superpixel segmentation, which uses a translation window for non-local mean (NLM) denoising algorithms in textured areas with rich changes However, these star map denoising methods have high requirements on parameter dependence, and there is still room for further improvement in denoising performance.

[0005] 2) Threshold segmentation method, which uses the mean value of the gray value of each pixel in the star map plus n times the error as the global threshold to separate the star target from the background noise, where n generally takes a value of 3 to 5, but the complex background is processed uniformly Noisy star map images often make it difficult to eliminate the band noise on the edge, and the processing efficiency is low

Aiming at the problems that the traditional wavelet threshold function is discontinuous at the threshold, and the wavelet estimated coefficients are deviated from the actual coefficients, resulting in distortion of the image after denoising and pseudo-Gibbs, some scholars have proposed a new improved threshold algorithm. In addition, Shi Chunlin and others also proposed that the one-dimensional maximum entropy method can be used to threshold the star map, but there is no global threshold method suitable for all star map images.

[0006] 3) Static background noise method, this method is based on the relative stability of the background noise when the star sensor images space, select two frames of star map images in the same orbit star map at intervals, and use the median value of the same pixel in the star map as the background noise, so as to achieve the segmentation of star targets and background noise, but the complex space environment makes the background noise of different periods of time different, resulting in the star map processed by the static background noise method will retain some system background noise, and the generalization is poor

Images