FPGA-based accurate star segmentation with moon interference

Star sensors, which are based on matching obtained star information to the star catalogue, are instruments widely used to determine a spacecraft’s attitude in space. Thus, a highly accurate extraction of real-time star information is a major issue in star sensor designs. In this study, a novel field programmable gate array (FPGA)-based accurate star segmentation algorithm is proposed to satisfy real-time requirements. Windows with a star or its parts are found using a maximum filtering based local gradient and local gradient threshold that is adaptively calculated using the local mean. An adaptive threshold, which is based on local mean and local median, can be used to determine whether the center pixel of the window is a pixel of a star. The algorithm can properly segment bright and dark stars, and completely eliminate moon interference. A precision of <0.09 pixels can be maintained in images at different Gaussian noise levels. A parallel and pipeline architecture also utilized in FPGA implementation, and the processing time is 22.22 ms for a 2048 × 2048 gray-level image.