Design And Implementation Of Video Mosaics System Based On Embedded And SURF Algorithm

Video image splicing technology is a kind of technology that to extract the characteristics of multiple images that have the overlapping areas, to match the images according to the extracted features again, to find out the most suitable transformation model, then to transform the matching images using the transform model, finally, to merge those images that have the overlapping areas together according to a certain rule of image fusion, resulting in a complete composite image with high resolution and large visual angle. At present, video image splicing is an indispensable research direction in the field of digital image processing, it have extremely extensive application in such aspects as geographical surveying and mapping, the aerospace, daily life, medical diagnosis and so on, and, with the continuous development of society, video image splicing technology will also be promoted in other emerging fields.Based on a large number of video image stitching technology research, this paper referred to a large number of documents, firstly introduced the background and significance of video splicing technology, analyzed the research status of video image stitching, looked forward to its broad application prospects, proposed the preliminary ideas to complete the video image stitching algorithms on embedded devices, and gave the structure of the contents of this paper generally. Then introduce the concept of embedded systems, including embedded hardware platform, embedded peripherals and embedded operating system, on this basis, select the embedded hardware devices of this article, by cross-compiling and transplanting the Linux kernel and bootloader files, establishing the minimal root file system, this paper constructs a minimum of embedded software systems. Then, the paper analyzes the general steps of video stitching algorithm, introduces the related knowledge about the characteristics of the image, image matching and image fusion, analyzes the key technologies, on this basis, through the analysis of the current video stitching algorithm, video stitching algorithm is proposed in this paper, which is based on SURF and improved RANSAC video stitching algorithm. Then details the specific steps of SURF algorithm, but also introduces the RANSAC algorithm specific practices(ie the improved RANSAC algorithm), as well as details the image fusion approach used in this paper(ie fade out the weighted average method). Next, on the hardware platform builded in front of the paper, this article completes the transplant and testing work about the proposed method, including the specific construction of embedded Qt + OpenCV environment in Ubuntn of virtual machine, the design of graphical user interface in the Qt environment, the design and writing of code in Qt, the related cross-compiling and transplantation work about the completed code and so on. After the completion of the work, then this article is debugging and testing for the algorithms, which is divided into two parts, one is to debug this algorithm on the PC, to test and document experimental data and analyzing those data; the other part is running on the actual algorithm in the embedded hardware by this article, to test and record the relevant data, and analyzing those data.Experimental results show that the algorithm that is running on embedded devices and on the PC, are able to obtain a better image fusion, but also basically meet real-time requirements. Therefore, the methods described here has a certain reference value in the research and application of video splicing technology.