Single-particle image spot on star map of star sensor can be mistaken for real star image and incur fault star identification, how to treat space single-particle properly in a star sensor is an emerging serious problem. Analysis and modeling are conducted for two types of single-particle image which are of spot shape and stripe shape. By observing two extreme cases of center location (pixel center or corner vertex) of a Gaussian star image and after examining of their Gaussian gray distribution characteristics, the discriminant criterion of single-particle image spot is established with a form of gray interval which is different with the pixel number of connected domains. Another discriminant criterion is also set up according to morphology, which is a limit value calculated from an equation expressing the slenderness of a star image. Simulating test results indicate that the involvement of the discriminant criterion can effectively identify single-particle images from real star ones, which helps to promote the reliability of attitude determination of star sensor.