Compared with other landing methods， visual autonomous landing has the advantages of obviously independent and simple system structure， and rich measurement information. The non-cooperative visual landing has great significance， but it also has higher requirements for the accuracy of feature extraction. In order to realize the stable and high-precision matching of non-cooperative feature points of reference location in the aircraft carrier image， this paper proposes a structural feature detecting and matching method based on straight line detection of line segment detector （LSD）to achieve accurate recognition of structural corner points on the aircraft carrier surface. Experiments show that compared with oriented FAST and rotated BRIEF （ORB）， speeded up robust feature （SURF） and other image features， the proposed method can locate specified structural features more accurately and thus can improve the accuracy of visual positioning effectively.