This paper presents a corner matching algorithm based on feature optical flow.Firstly，the homogenized corners are extracted.Secondly，judging from the improved approach of feature optical flow estimation，the local optical flow estimation on the corners can be calculated，rough optical flow vectors are gained and approximate matching of the corners is achieved.Finally，optical flow estimation of corners is acted as constraint condition.According to the coordinate relations between two cornersets of images，abnormal corners are excluded.In this way，the corners of the two images can be accurately matched.The simulation experiment shows that the presented algorithm can reduce the computational complexity and raise matching precision so as to precisely track moving object.