Fast Matching System Design Based On SIFT Algorithm

With the development of science and technology, Image matching is playing amore and more important role in human’s production and life. Image matching is amethod based on image content, structure, characteristics, texture, relationship and gray,consistency, and similarity analysis, to seek similar image target quickly. Imagematching technology has a very wide range of applications, such as: the industrial test,aircraft navigation, the terrain matching, the guidance of the projection system andmeteorological forecast, resource analysis, traffic management, medical diagnosis,character recognition, etc. Image matching plays a core role in various areas.The article is based on the SIFT and KD_TREE, first proposed the concept ofimage fast matching system, and has carried on the detailed system design and modulepartition, it has realized the system function design and the logical partition, at the sametime, it also has realized the synchronous matching for more pictures. The study ofimage fast matching is of great significance.Scale-invariant feature transform (SIFT) algorithm was proposed in1999, andwas perfected summary in2004. SIFT algorithm keeping invariance for image scaleand rotation. To a certain extent, it also maintains the stability of perspectivetransformation and noise. So, it can do a good job on image feature point extraction.Research focused more on the single image feature points extraction algorithm inthe related research, Such as SIFT, SUSAN, Harris and GLOH, SURF and otherimproved algorithm based on SIFT feature point extraction. Even in some papers, theresearch about feature point matching was also put into the SIFT algorithm. In thispaper, a feature vector in an image using KD_TREE access image fast matchingmethod based on the SIFT algorithm was proposed. Design and implement imagefeature point extraction and matching function modules. At the same time, do thedetailed design and illustration based on the whole system implementation process.Implement the multiple pictures of synchronous matching function, largely reduce thesystem overhead, improve the matching efficiency of the system.The system was divided into four function modules: Feature Vector ParameterManage (FVPM), Feature Vector Data Manage (FVDM), Feature Vector ExtractAlgorithm (FVEA), Feature Vector Match Algorithm (FVMA). FVPM is mainly responsible for system parameter setting, FVDM is mainly responsible for accessfeature vector, FVEA is mainly responsible for extract the feature points in the image,FVMA is mainly responsible for matching image feature vector. On the one hand,system achieves the function of the source folder learning, on the other hand,implement treat matching image matching function. In the end, the system was tested,and analyzed, the results shows that function of the system design is basically realized.