Study On Image Mosaic Technology And Its Application In Railway Video Monitoring System

With the diversifying development of Image Mosaic Technology in modern, more and more algorithms have been worked out; most of them were based on the method of feature point's registration. The mosaic of feature matching is what can figure out the image transformation parameters by matching features of some partially overlapped images in the first place, and then transform projection of images and mosaic images by different fusion methods. Image Mosaic Technology has been widely used in applications, such as camera panorama mosaic, medical image registration and mosaic, fingerprint identification and mosaic, and so forth.Based on feature point matching aspect of image mosaic, this paper firstly aims to explain image coordinate transformations and fusion methods on the basis of introductions about essential factors and main methods of image registration. Secondly, follow feature point as a main line and summarize principles and properties of each corner point as a result of some experiments, such as abstracted experiment, matched experiment and eliminated dispatch experiment. Finally, bring SURF and SIFT algorithms in research and verify the stable scale and rotation of them, which can achieve image mosaic further. The method of image coordinate transformation is called perspective transformation during the process of mosaic. The image method of smoothness, transition and fusion is called gradually fade-in or fade-out method. In the aspect of reality, we can focus on the problem of railway surveillance system about limited scopes which is caused by using only one surveillance camera system. So this paper will explain how to put image mosaic technology into railway surveillance system. To begin with, picture the images by using two surveillance cameras. Moreover, match these partially overlapped images quickly by using SURF algorithm. Finally, complete the images and show them in the same interface. To a certain degree, this method expands the surveillance scopes.This paper mainly carries out an algorithm of inputting images and ordering them automatically in the process of mosaicing because of stable scale, rotation and illumination of SURF and SIFTS algorithms. This algorithm doesn't influence mosaic result at all, which can ensure not only the accuracy of image registration for a greater degree, but also correct mosaic result even under the circumstance of reversal interface. In a word, it further improves system's reliability. What's more, the fast mosaic of surveillance video by using SURF algorithm also has great definition and stability.。...