The SIFT feature matching can overcome the influence of revolve and scale. It can recognize runway effectively under the complex flight environment, but its computation is heavy. In order to reduce the computation, this paper proposes a new fast method of SIFT feature extraction. This method uses a group of concentric squares to construct the Pyramid-feature-descrptor, calculates each seed vector of the Pyramid-feature-descrptor by using the recursion algorithm, and carries simple sort on the elements of each seed vector to maintain the revolving invariability. The experimental result of identifing the runway indicates that this algorithm not only can increase the speed, but also can identify the runway effectively in revolving and scale.