Visualization Of Underground Structures In Augmented Reality

Underground structures are one of the most important infrastructures of the cities. Their security operating is a prerequisite of the high efficiency and high quality for modern cities. The traditional management can not display the real world scene and the virtual objects in the same screen. The user is not able to register the virtual data with their real working environment, which limits the applications of these systems.Augmented Reality (AR) is a computer technique that integrates digital information and real videos into a single display at interactive rates. Such display can enhance the user's perception of the real environment by showing information the user can not directly sense when unaided, thus AR is widely used in medical, architectural, exploratory and military applications. With real-time camera tracking, AR system can directly display the virtual pipelines on the scene image, and provides immersive experience for the constructors. Simple overlay of rendered images can reveal the accurate space information, but produces a confused illusion that the virtual pipelines are floating on the ground. Furthermore, it can not overcome the perceptual issue of depth ambiguity in pure virtual visualization.This paper presents an on-line AR visualization system of underground structures. Our system transcodes geographic data into rendering-friendly intermediate data that is then used to generate the 3D models. There are two primary perceptual issues involved with obscured information visualization in AR:1) perception of occlusion and depth,2) preserving and emphasizing important information. To address them, we propose an X-Ray technique based on Focus & Context theory and a visualization method based on vertical clip planes. The former improves the traditional X-Ray in AR by extracting features from both video image and 3D virtual objects, and then uses the features to provide visual cues for depth perception and supply sufficient information transfer by adaptively removing redundant details. Inspired by section view in Engineering Graphics, we propose a vertical clip plane based visualization technique. It allows the user to place the clip planes, and then controls the rendering styles and composition of virtual and real objects according to the viewpoint, position of clip planes, video and 3D objects. Visualization of virtual objects with preserved scene context precisely exhibits the depth information of the underground structures surrounding the clipping planes. This method improves visual accuracy of virtual information, and enhances correctness of occlusion perception among the objects. Both of our approaches do not require 3D model of the physical scene, and provide real time performance in modern graphics card.