关键词: 不变矩, 16位灰度图像, 小目标检测, 空间图像

Abstract: Registration and difference can’t exclude stars when processing the 16-bit gray images because interpolation error of high light star is larger than the difference of target and background.Focus on this problem，a detection method of space small targets is presented based on moment invariant which is a concept of pattern recognition.This method starts from relative position of stars and avoids from removing stars by gray difference.It finds three stars as reference star group used the character that star moment invariants are equal in neighbor frames.The unknown points are joined to reference star group one by one and stars are extracted by judging whether the moment invariants of group are still equal after inserting new point.Targets are extracted from residual points using the continuity and consistency of targets movement.The problem is illustrated by registation and difference of emulation stars.The contrast experiments between registation and difference method and moment invariants method used real sequence images are processed.Results indicate that stars can be wiped off by this method which can’t be removed by registation and difference and the method can detect small targets from 16-bit gray images efficiently.

Key words: moment invariants, 16-bit gray image, small targets detection, space images