基于 FPGA 的局部天区星图识别技术

摘要

为了提高局部天区星图识别的实时性和抗噪声能力，提出基于 FPGA 平台的 SOFM 神经网络星图识别方法。首先从局部天区内选取导航星、构建导航三角形，然后利用 SOFM 网络的聚类功能对导航星三角进行归类，最后将训练好的网络应用到 FPGA 上，充分发挥 SOFM 并行性算法的优势，以并行流水线的方式来星图识别。仿真结果显示：SOFM 算法比传统的三角形算法具有更强的抗噪能力，而使用 FPGA 进行识别比使用串行处理器在速度上可以提高20多倍。%In order to enhance the level of real-time and anti-noise ability in local sky star map identification, a star map identification method based on FPGA is put forward which uses of SOFM network. Firstly, navigation triangles are built with navigation stars selected from the local sky. Then, the triangles are classified by the clustering ability of SOFM network. Finally, the SOFM network is applied to FPGA after it is well trained, so the star map can be identified in a parallel flow way. A simulation is conducted and the result shows that the SOFM star map identification method has a more powerful anti-noise capacity, and its speed is over 20 times more than a serial processor.