Research On Feature - Based SIFT Image Matching Method

Currently, the digital image processing technology has been rapid development, the image matching is one of the key technologies, has been in the aviation image mapping, missile terminal guidance, medical image processing, computer vision, optical radar automatic search, remote sensing data fusion, automatic docking of the 3D image reconstruction, pattern recognition, change detection, space station and many other fields get attention and application.Due to the difference of image matching, the matching method is very different, and there are some advantages and disadvantages. Image matching based on the characteristics is an important method of image matching, compared with the other matching method is to consider various factors, for example to match the brightness of the image area, the texture changes. Reasonable and effective use of feature-based image matching can effectively achieve the registration of multiple images. This paper mainly discusses the following aspects:This paper discusses the basic concept of image matching method based on feature, and expounds the principle and method of using this technology. Two kinds of commonly used extraction methods are introduced in this paper which are the point feature extraction and line feature extraction. And according to the different algorithms of the two extraction methods, design related experiment. Finally, according to the experimental results, a detailed contrastive study of the differences between different algorithms, analyzes the advantages and disadvantages of different algorithms.SIFT algorithm is widely used in image matching, this paper on the SIFT algorithm in four different data sources to match the image of the experiment, focusing on the analysis of SIFT algorithm in different data source image matching effect. By modifying the SIFT algorithm for image matching threshold ratio, respectively for different sources of data matching experiments, statistical analysis of matching reference factors, study optimal proportion and threshold of each source image data in the SIFT algorithm, sift operator in different data sources of matching test results obtained.Experimental results show that the SIFT algorithm has good extraction effect, and is not affected by the change of image radiation effect; the number of threshold ratio adjustment and feature point extraction, image matching accuracy exist linear correlation.