Research On Target Recognition Algorithm SIFT Technology

In recent years, automatic target recognition technology is used in military, industry and agriculture more and more widely, and the research of this technology has already been a hot topic in international academic research. The basic theory of target recognition is to estimate various physical characteristic parameters through extracting the feature information and recognize the target based on the discriminant function at last. The key to target recognition is to extract the feature of target and determine the category of target based on the extracted feature information. Therefore, the emphasis of this paper is to extract and recognize the feature of target. At the present time, there are many kinds of algorithms to extract and recognize the feature of target in the area of target recognition at home and abroad, but any of them has its own characteristic and sphere of application. It’s necessary to select and improve the algorithms properly based on the surroundings and the purpose of research.After analyzing the relative literature both home and abroad, this paper is to adopt the SIFT algorithm to extract and match the feature of targets based on the concrete purpose. And all of the algorithms in this paper have been emulated on MATLAB.The SIFT algorithm is the short of Scale-invariant feature transform, it not only has good invariance to the change of scale, illumination and space rotation, but also can extract huge amounts of feature information from a small number of targets. So the algorithm is used in the image matching widely. Although the SIFT algorithm has good capability and prominent merit, there must be the mismatches inevitably in the match result because of the influence of kinds of objective factors. Therefore, this paper paper proposes a new theory to reduce the mismatches-the theory to reduce the mismatches based on the main orientation of keypoints. It has been proved that this method can reduce the mismatches effectively and improve the accuracy of SIFT algorithm greatly through the simulation.