一种星辐射模式下全天区星图识别算法及FPGA实现

A Star Identification Algorithm for All-Sky Regions in Star Radiation Mode and Its FPGA Implementation

摘要:

针对在DSP中实现全天区星图识别算法耗时长导致实时性差,提出并实现了一种基于FPGA的星辐射模式下的全天区星图识别算法。该方法利用FPGA并行性、高主频等特性,在保证算法高精度、高准确度的前提下,解决了全天区星图识别耗时长的问题。首先,根据星点信息构建星辐射模式；其次,基于FPGA实现了星辐射模式下对应特征向量的计算；最后,提出了一种可预读、流水化的星辐射模式匹配算法,极大地缩短了星辐射模式匹配的时间。在Xilinx公司的Xc7a75tfgg484-2平台上进行实验验证,结果表明,在相同时钟主频下,基于FPGA的星图识别算法速度比基于DSP的算法速度快10.95倍以上,而且在使用三组特征向量同时进行匹配时,识别率超过99%。

Abstract:

Aiming at the problem of poor real-time performance due to the time-consuming implementation of the star identification algorithm for the all-sky regions in DSP, a star identification algorithm for all-sky regions based on FPGA in star radiation mode is proposed and implemented. This method took advantage of the characteristics of parallelism and high main frequency of FPGA, and solved the problem of long time-consuming of star identification in the all-sky regions under the premise of ensuring high precision and high accuracy of the algorithm. The main works of this paper were as follows. Firstly, the star radiation pattern was constructed according to the star point information. Secondly, the calculation of the corresponding eigenvectors in the star radiation pattern was realized based on FPGA. Finally, a pre-readable and pipelined star radiation pattern matching algorithm was proposed, which greatly shortened the time of star radiation pattern matching. The experimental verification was carried out on the Xc7a75tfgg484-2 platform of Xilinx Company. The results show that under the same main clock frequency, the speed of the star identification algorithm based on FPGA is more than 10.95 times faster than that based on DSP. And the recognition rate exceeds 99% when using three sets of eigenvectors for simultaneous matching.