3D Indoor Scene Reconstruction With The Active Omni-directional Vision Sensor

Three-dimensional reconstruction of the scene plays a very important role in many areas. For existing methods, acquiring geometry and color information of the objects simultaneously is difficult. Moreover, the current techniques can not store, modify and model the point clouds efficiently. In this work, we have developed a novel sensor system, which is called active stereo omni-directional vision sensor(ASODVS), to address those problems. It is more convenient and faster to construct mesh model of the 3D points cloud obtained with ASODVS considering the ordered feature of the data. The reconstruction result is more realistic with the display method developed here called observer-centered 3D drawing. Experimental results demonstrate that the proposed ASODVS system can measure and reconstruct the 3D space in real-time and with high quality. The main contents of this thesis are as follows.1. Research on the Active Stereo Omni-Directional Vision Sensor(AODVS). It is an integrated system composed of a single view Omni-Directional Vision Sensor(ODVS) and a mobile planar green laser generator platform. Driven by a stepper motor, the laser platform can move vertically along the axis of the ASODVS and emit laser in 360 degree of the horizontal plane. The ODVS captures panorama pictures with laser of the scene in real-time. The geometry information and RGB values of points in the space can be calculated by parsing the panorama images with a software developed by ourselves.2. Research on the 3D reconstruction of the indoor scene. The 3D points model is stored as a matrix in the software system based on the ordered points cloud captured by ASODVS rather than scattered points cloud. The 3D mesh model is acquired based on features of the points cloud data without constructing topological structure.3. Research on 3D drawing from the observer. It is necessary to distinguish the displaying of objects and scenes considering the particularity of scenes. 3D drawing from the observer developed in this thesis is useful to display models more realistically and make observers feel more immersive.4. The software implementing of indoor scene 3D reconstruction. The whole software system for scanning, reconstructing and displaying of indoor scene is realized with Java and Java3 D. It is able to capture points cloud data, construct mesh model and display three-dimensional model from the perspective of observer.