Image Matching Technology Based On Asift Characteristics Research

With the continuous development of image sensor technology in recent years, Image matching has become a research focus in many areas, such as image and scene understanding, target recollection and tracking, computer vision and other crucial areas. Image sticking is an image processing technology detecting the degree of similarity of different pictures, which are of different perspective, of different sensor and obtain under different photograph conditions. Not only that, it also includes pictures of the same sensor taken in different times.In recent years, with the continuous improvement of the image sensor, image matching has become a hot research topic in computer researching areas such as image understanding, computer vision as well as fields of remote sensing. In medical and biological fields, we can’t get a complete picture of the samples with the viewing angle changes and different magnification times of microscope. So, it’s a research puzzle to match and identify microscope images. This paper focus applications of image matching based on modified ASIFT algorithm. The modified ASIFT algorithm can be applied to solving image matching. The main work of this paper can be summarized as the following aspects.First, this paper summarizes image matching technology, then analysis research significance, status and application areas. At last, this paper proposes an image matching method based on the optimized ASIFT algorithm. The improved ASIFT algorithm solving cluster phenomenon of ASIFT feature points according to establishing anti-affine camera model. The specific steps are as follows:Using optimized detection process to find out all ASIFT feature points in all scale spaces, and then eliminate cluster phenomenon with a circular template detector. At last, in matching process, this paper uses improved RANSAC algorithm to purify matching pairs for a better precision. Experiments results show that this optimized ASIFT algorithm can solve phenomenon caused by rotation of the angels. As a result, the algorithm proposed by this paper can find out more valid feature points. This modified algorithm optimizes flow of the original ASIFT algorithm, fully digs up parallelism of this algorithm. This research does lots of experiments based on pictures of different sizes and that taken under different angles. An experimental result validates this improved algorithm in image matching and image mosaic.