Aircraft Model Identification Method Based On Invariant Moments Research And Realization

Aerial images in the modern military in the field of application of more and more widely, more and more people started to how to conduct aerial images fast and accurate identification of some concern. In particular, rapid identification of those aerial imagery of the tactical objective, which in modern military warfare played an important role. In a variety of targets, due to the aircraft model is very large, and a variety of different types of aircraft, the performance gap between the relatively large, so the aircraft tactical objectives of this analysis would not only need to determine whether there are planes, but also the need for able to accurately identify the aircraft type.As information technology and computer technology development, especially the development of pattern recognition, aircraft recognition became a popular pattern recognition of the branch. How to use the computer to detect targets on aerial photographs, the implementation of automatic identification, intelligent judgments target type, not only can improve the detection system, the degree of automation, to enhance the Air Force aerial reconnaissance intelligence capabilities also have an important role.The main research contents of this paper is how to identify aircraft targets using moment method. The main findings include: for discrete cases, Hu invariant moment features only the translation, rotation invariance, the proportion of invariance does not have the problem, suggestions to improve; through the collection of various models of aircraft vertical view information on the use of PhotoShop is processed into black and white 2 binary image, using models extracted image feature recognition system to construct the image characteristics of the sample database; used 6 Hu invariant moments, four affine moment invariants, boundary complexity, aspect ratio, etc., as identifying characteristics, To identify targets for the use of a tree classifier methodology and the implementation of the template image discrimination.Theoretical analysis and simulation of knot-ming, the proposed algorithm is feasible. After the amendment, the recognition rate has improved significantly.