The extraction of navigation stars is a critical step in near-infrared celestial navigation. It is related to the subsequent star target identification and star image matching directly. Because of the strong background of the sky, the extraction of star targets is extremely difficult. If the near-infrared celestial navigation is used to detect stars in the near-infrared waveband, enough star targets for navigation can be acquired even in the daytime. On the basis of analysis of the characteristics of near-infrared star images, an energy projection method is introduced and the potential star targets and noises are located according to the peak point of the projection. Then, the areas to be processed are set up according to the energy variation. The targets and noise points are extracted quickly by suing the OTSU algorithm. Finally, after the multi-frame overlay is used to remove the noise points, the star targets are extracted. The experimental result shows that this algorithm is simple in calculation, short in operation time and higher in detection ratio. It can detect the star targets against a complex background effectively with a low false alarm rate.