Design And Implementation Of A Multi-channel Real-time Panoramic Video Stiching System

In order to solve the problem of insufficient width of the field of view camera when you get single scene existed, video splicing technology was put forward. This technique can be used to display the target object around the scene without blind zone. and then to fill a wide range of site-specific, no blind spot monitoring market vacancy.We can foresee, panoramic video display system will be widely used in the vehicle(ship) overload monitoring, security surveillance in public places, intelligent transportation, panoramic mapping and other military and civilian fields.This paper introduces an improved panoramic video splicing system architectures, this architecture is used to solve real-time video splicing, since operation efficiency and improve resource utilization, has a natural advantage, and analyzes this architecture hardware platform and environment to build process.Then with the core video splicing process steps, some common algorithms are proposed, such as the use based on the initial value or based on the feature point method for image registration, based on the optimal use to suture method or method based on Laplace pyramid image fusion, but also,the different environmental advantages and disadvantages between them were analyzed.Next comes the design and implementation of the entire video display system, including the design and optimization of the DM8168 end video stitching algorithm, From PCIe driven design to the final realization of panoramic video of a specific generation. In optimizing video stitching algorithm, this paper respectively timeliness, accuracy of registration and other issues, Some corresponding parallax approaches were put forward. Match uses a multi-resolution image fusion method based on optimal suture solve parallax generated; an improved based on SURF(Speeded Up Robust Feature) feature point extraction algorithm, data processing speed and improved video registration accuracy problem. PCIe is designed to accommodate a large number of real-time transmission of video data, EDMA was used to handling data,it can improve the data transmission speed to achieve real-time processing of image data requirements..Finally, Some corresponding experiment were made according to the optimization algorithms and architecture design. Compared with the spliced and original images, the delay between them is very low. which can be maintained within the range of synchronization. At the same time, after the entire video is mosaic, it is displayed smoothly, and it can satisfy the market surveillance requirements.