Abstract: To improve the angular resolution and data update rate of star trackers, high-speed, large-area, multi-channel output image sensors have been widely used. The rate of traditional centroid extraction algorithms has become a major obstacle to the data update rate. To address this issue, this paper proposes a multi-channel high-speed star extraction method based on the all programmable system-on-chip (ZYNQ). This algorithm effectively leverages the parallel processing advantages of the field-programmable gate array (FPGA) side of ZYNQ and the serial processing advantages of the advanced RISC machine (ARM) side. First, on the FPGA side, star field data from multiple channels are simultaneously processed for identification, extracting the necessary star point information for calculating centroids. Then, on the ARM side, the star point information is used to merge star points across boundaries, and finally, the centroids of all star points in the star field are obtained. This algorithm enhances the star point centroid extraction rate and solves the problem of merging star points across boundary. Finally, experiments on the hardware platform show that with an input image resolution of 2048 \times 2048 pixels, the maximum frame rate reaches 301 fps, validating the correctness of the algorithm.The highest frame rate of this algorithm is 110% higher than the reference 15 algorithm, and it has important application value in improving the data update rate of star sensor.