Abstract: Video surveillance system is one of the most important tools for security protection. In this article, the multi-channel video sequence is obtained by decoding the video stream, and then the video texture mapping is completed by estimating the pose of video camera through the registration of the 3D model and the video texture. Moreover, texture fusion technology is used to eliminate the color difference seams between each channel and finally the real-time projection is achieved. The experiment result illuminates the advantage of this arithmetic and the idea of rendering the live video on 3D model. The idea of rendering the multi-channel video on three-dimensional scene can make it convenient to record the real images and grasp the real-time monitoring of the scene information for monitoring site security, which provides the basis for safety protection and emergency disaster warning and rescue.

Key words: video stream, 3D scene, augmented virtual environment, video surveillance system, texture projection, video registration, texture fusion