Abstract—In this paper, a new star recognition method based on Delaunay Triangulation (DT) algorithm and distributed neural networks was proposed to decrease the search space and increase the star recognition success rate for star sensors. It computed the DT of all stars in catalog. Then, it employed this method on stars in the captured image. It compared the generated DT graph of stars in the image with the catalog graph by using Relaxation By Elimination (RBE) method. RBE used an ‘inverted’ relaxation labeling method that found a good match of the input graph with the catalog graph. RBE was implemented by Correlation Matrix Memories (CMM). CMM was a kind of neural networks to store the constraints between the nodes of the graphs being searched. This algorithm relied on the positional relations and angular distance of stars in each triangle that is made by DT. The experimental results showed that when the position error was about 120 arc seconds, the identification success rate of this method was 89% while the identification method based on the matching probability was only 77%. In addition, the storage requirement of the algorithm was small.
Index Terms—Star identification, delaunay triangulation, graph matching, relaxation by elimination.
The authors are with the Amirkabir University, Tehran, Iran (e-mail: firstname.lastname@example.org, email@example.com).
Cite: Saeideh Sadat Miri and Mohammad Ebrahim Shiri, "Star Identification Using Delaunay Triangulation and Distributed Neural Networks," International Journal of Modeling and Optimization vol. 2, no. 3, pp. 234-238, 2012.