Passive infrared(IR) detecting technology has been more and more appreciated in military domain because of its unbeatable advantage. Dim and small moving targets detection of IR sequence images is a key technique for the IR detection and tracking, alarming, guiding system. This thesis mainly deals with the research on the methods of dim and small moving targets detection of IR sequence images,and one of the methodology flow is presented.Firstly, since the IR sequence images are not staring scence, to facilitate the following treatment, some registration algorithms on the IR sequence images are discussed. According to the advantage and disadvantage of these registration algorithms and the fact of "dim" "small" target, classic registration algorithm----block match is chose, and a method of self-adaptating template selection for block match is presented. Secondly, on the basis of the analysis of the spatial-temporal feature of background and infrared targets in infrared sequence, a method for clutter suppression and noise removement based on the association of spatial self-adaptating Two Dimensional Least Mean Square(TDLMS) background predict and temperal Higher-Order Cumulant(HOC) is presented, and its improved version suitable for strong edge rejection and gaussian noise suppression is proposed. Finally, according to the continuity and regularity of target moving, a target searching method based on spatial pipeline cumulation and temporal neighborhood judgement is presented, and a target tracking method based on the model of track memory is discussed, introducing the Kalman tracking filter for long-time stable tracking. |