Parallel Pipeline Algorithm of Real Time Star Map Preprocessing

To improve the preprocessing speed of star map and reduce the resource consumption of embedded system of star tracker, a parallel pipeline real-time preprocessing algorithm was presented. The two characteristics of the mean and the standard deviation of the background noise of a star map, were firstly estimated dynamically, and the influence of the star spot itself on the background noise characteristics was removed. The criterion for deciding whether the following noise filtering is conducted was established, and further the extracting threshold value which depends on the level of background noise was assigned, and meanwhile the centroiding accuracy was garanteed. In the processing algorithm, as low as two lines of pixel data were buffered, and only 100 shift registers were needed to record the connected domain label, by which the problems of resources wasting and connected domain overflow were solved. Simulation results show that the necessary data of the selected bright stars can be immediately accessed in a delay time as short as 10 μs after the pipeline processing of a 496×496 star map in 50 Mb/s is finished, and the needed memory and registers resource total less than 80 kB. To verify the accuracy performance of the proposed algorithm, different levels of background noise were added to the processed ideal star map, and the statistic centroiding error was smaller than 1/23 pixel under the condition that the signal to noise ratio was greater than 1. Therefore the validity of the parallel pipeline algorithm of real time star map preprocessing is verified, which helps to increase the data output speed and the anti-dynamic performance of star tracker.