Research On Fast Reconstruction Of Scene Based On Asodvs

After years of development, there have been some achievements in the area of stereo vision. However, the feature point matching is still the most difficult part of stereo vision; Active stereo vision can be better to obtain the depth information of the scene without the process of point matching problem, but often fails to provide the panoramic depth of the scene. In order to achieve the depth of the panoramic scene, we propose an active three-dimensional panoramic sensing technology. Our work contains three aspects:1. The study of Active3D Omni-directional Visual Sensor. Based on line laser, we designed a new laser source which can generate line laser around the scene. Two of our designed laser source were combined with panoramic vision sensor to form the proposed Active3D Omni-directional Visual Sensor.2. The study of depth measurement of the scene. In order to calculate the depth of the laser feature points, it is necessary to identify and extract the laser pixels in the Omni-directional image. Based on the RGB color space and the HSI color space, we proposed a way to locate the laser pixels. According to the triangular relationship of the geometry, the depth of the laser points were calculated. Moreover, the three-dimensional coordinates of laser points also can be obtained. Experiment results show that the depth calculated have a certain degree of accuracy.3. The study of fast three-dimensional reconstruction of the indoor environment. Bird-view transformation of Omni-directional image was used to get the texture of the ground scene. Here, we use three different ways to achieve the Bird-view transformation of Omni-directional image. Combined depth of laser points with the proposed multi-plane geometry model of the indoor scene, we reconstruct indoor scenes. Experiment results show the proposed multi-plane geometry model can reflect to the real indoor scenes.