Real-time 3D Reconstruction Using Self-adaptive Structured Light

Recently, 3D measurement technology widely attracted the industrial automation's attention. A lot of 3D measurement equipments have been developed, such as on-the-shelf 3D scanner, laser range finder, stereo camera, BGA measurement and so on. The object 3D measurement is an important orientation in computer vision. It is used to fetch the 3D shape information of the object. There are many 3D measurement methods now, including stereo, structured-light measurement, probe measurement, reconstruction based on texture, reconstruction based on light and shade, reconstruction based figure, etc. Each method has diverse advantage, disadvantage, and scope of application. Structured-light is widely used for its simplicity, precision and unbroken.Structured-light is regular and .organized light. The structured-light 3D measurement method could be described as follow: use structured-light pattern to project on the object, and then translate the 2D feature-points on image to 3D feature-points, at last rebuild the shape of object by these 3D feature-points.The main content of this paper is about the research of real time adaptive structured-light modeling method, which includes 3 parts:First, the decoding of structured-light pattern, which should detect as more feature-points in the images as possible, is useful for pattern coding and decoding. In this paper, all kinds of coding and decoding method is introduced, and also is a method that improves the adaptive structured-light decoding speed and precision.Second, the analysis of structured-light 3D information is the method that calculates the 3D feature-points from 2D feature-points on images. This paper introduces the traditional 3D reconstruction method based on triangulation. And then it provides a novel 3D reconstruction method based on homograph matrix. At last it compares with these two methods and points the advantage of new method in speed.Third, 3D reconstruction is the method using 3D information of feature-points to get the shape of the object. It finishes the reverse engineering work, which builds the modes by connecting the point-clouds from structured-light method to triangle facets, and defines the normal vectors to each facet. Considering the complexity, it would take much time when reconstructing the dispersed point-clouds. This paper provides a faster method using the feature-points information from pattern.At last, it shows the results and analyzes the experiments data. Besides the work of real time model by adaptive structured-light, there are many work need to be done in the future. The paper gives the goals for the future work.The innovation of the paper: the improvement of decode structured-light 2D information algorithm by adaptive method, the improvement of decode structured-light 3D information algorithm and the improvement of reconstruction algorithm by coding and accelerating divided mesh.