When using LOS guidance weapons, the missile cursor signal often shilds on target image, the target is prone to be lost and that would influence reliability and accuracy of target tracking. The paper applies threshold segmentation to seperate missile tail-flame from target first. Secondly, corner detection was used to acquire object features and track the target, and then corner matching was used to obtain the corresponding relation of cornes in two adjacent fields. At last, affine transformation was applied to coordinate transformation with the position of pixels in the fill area. Finally, the dynamic object which is shielded by missile cursor signal can be reconstructed. Simulation results show that the algorithm can effectively renew target image in real time and improve target tracking accuracy.