一种采用行列聚类并行的星像坐标提取方法

An Star Coordinates Obtaining Method Based on Row and Column Clustering in Parallel

关键词:  星敏感器  星像坐标  图像处理

Key words:  Star sensor  Star coordinate  Image processing

摘要:

为了提高星敏感器的数据更新率，必须减少星像坐标的提取时间。在提取星像坐标的过程中，现有基于FPGA的方法虽然避免了星图保存和读取过程，然而需要高性能的FPGA才能实现。为此提出了一种采用行列聚类并行的星像坐标提取方法，将逐个像素聚类的方式优化成像素队列聚类的方式，实现了对恒星星像坐标的实时提取，可以解决现有的实时提取星像坐标方法中像素输出时钟与处理时钟不匹配的问题，达到1个时钟周期处理完1个像素的目的，从而减少了星像坐标的提取时间，提高了星敏感器的实时性。最后把该方法嵌入到某星敏感器系统中，实验结果表明，与现有基于FPGA的实时提取星像坐标方法相比，在相同工作频率且不降低星像坐标提取精度的前提下，提取时间仅为现有方法的21.1%，数据更新率提高了近5倍。因此，该方法不但具有实时性，而且具有较好的鲁棒性。

Abstract:

In order to improve the update rate of star sensor, the time of obtaining star coordinates should be reduced. The current method based on FPGA avoids the process of star image saving and reading，but the high-performance FPGA is needed. So a novel star coordinates obtaining method based on row and column clustering in parallel is presented, which optimizes the clustering method pixel by pixel into the clustering of pixel queues. The method can solve the existing problem that the pixel output clock does not match the processing clock with the obtaining star coordinates method. And the purpose of processing one pixel in each clock cycle is achieved, which results in the reduction of the star coordinates obtaining time and the improvement of the real-time performance of the star sensor. Finally, this method is embedded in a star sensor system. The experimental results show that on the premise of the same working frequency and without reducing the precision of star coordinates obtaining, the obtaining time is only 21.1% of that of the existing method, the update rate is increased by nearly 5 times. Therefore, this method is not only real-time, but also of good robustness.