A fast detection algorithm of cooperative object corners based on SIFT partition bidirectional matching

A challenging task of airborne operations remains the landing on the carrier deck, which limits the carrier operational efficiency during rough sea. Although existing corners detection algorithm can work well without consideration of scale and orientation distortion in cooperative object, extracting corners of cooperative object at the condition of scale distortion and orientation distortion even needs more to considerate. To solve this problem, this paper presents a novel approach for UAV's automatic landing on the ship's deck. We present the design of the cooperative object and then extract the corners of cooperative object based on Scale-invariant feature transform partition bidirectional matching. Firstly, SIFT features of cooperative object are extracted from the reference image and real-time image respectively. Based on this, the strategy of partition bidirectional matching is used to eliminate the error matched points. At last, the corners of cooperative object in real-time image can be obtained through calculating the affine transformation between reference image and the real-time image. Experimental results show that the proposed algorithm is not only accurate and robust, but also feasible for real-time applications.