Research On Image Sequence Stitching Algorithm And Its Application

Image stitching technology, an interdisciplinary subject,which involves in computer graphics, pattern recognition, image processing, and other aspects, has been a research focus in the image processing and computer vision. In the video understanding and analysis, target detection and tracking, image processing and analysis, etc have widely applications. Therefore, the study of image stitching technology is very meaningful.In this article, we take the Electro-Optical Distributed Aperture System on F35 and electro-optical tracking system as the main reference, taking the background of the project named integrated platform into account, and put forward the project of target tracking and image sequence stitching. Specifically, the scheme is mainly composed of two parts: one is the implementation of target tracking from double perspective image sequence; another is the implementation of real-time image sequence stitching from double perspective sensor, and marking the target use the tracking information.Based on the above introduction plan, this article studies image stitching technology theory and application, namely the second part of the scheme, and the first part of key research carried out by other members of the team. In the end, according to the requirements of the overall plan, we will integrate the two parts of research results. In this article, the main work is as follows:1、On the basis of analysis of existing research and the main processes involved in the mainstream image mosaicing algorithm, introduce in detail the image stitching involved in the main image imaging model, transform model and the basic principle of image processing. According to the practical application, to determine the overall algorithm based on the features of image matching framework.2、In view of the present mainstream image stitching process, considering the actual application requirements and scenarios, we put forward the method that by narrowing the scope of feature extraction method, achieved the purpose of reducing algorithm running time and improving the real-time algorithm.3、In the feature matching stage, put forward the method that according to the color at points, area color difference and the difference of regional structure, build the secondary feature vectors for the area around the initial extraction of feature points, then through secondary features matching, eliminate some mismatch in advance, the last stitching precision is improved.4、In this article, based on the actual application scenarios, the method that using target tracking information, changing the traditional linear fusion to non- linear fusion was improved. This method is better solved the actual scene ghost problem caused by a moving target.5、Under the Windows 7 system, vs2010 and opencv,implement the task that image sequence stitching and display the target in its result.