Research On Three-dimensional Scene Reconstruction Based On Structured Light

Three-dimensional scene reconstruction is one of the important directions in the field of computer vision research.It has been widely used in robot positioning,augmented reality,virtual reality,and autonomous driving.The 3D scene reconstruction method based on structured light,with its fast speed,high precision,strong robustness,non-contact property,has high research significance and practical value.In this paper,the key technologies such as line structured light and coded structured light are studied.New solutions are proposed for some problems,and an experimental verification is built on the basis of the proposed methods.The main researches are as follows:1.An offline reconstruction method based on binary space is proposed,to solve the shortcomings of traditional 3D scene reconstruction based on fringe boundary coding and decoding.In the traditional method,the code value of fringe boundary line is determined according to gray level of its neighborhood,and there are many decoding bits,which leading to a high boundary line error rate.In addition,the reconstruction process is prone to errors due to problems such as stripe diffusion and boundary disconnection.In this paper,based on the binocular structured,the binary coded stripe sequence is layered in the order of projection,and binary space is constructed to provide additional information for decoding.During the process of decoding,complete and accurate stripe boundary lines are extracted in the subspace according to the direction of black and white stripes in the binary space,and the code values of boundary lines are determined in the order of its extraction.The proposed method adds space and unique constraints to the extraction of fringe boundary line by using binary space,and considers the consistency of the states of corresponding binary spaces in the left and right perspective,so it can accurately match all fringe borders.The experimental results prove that the method has high accuracy.2.A real-time reconstruction method based on scanning is proposed,aiming at the problem of reconstructing large-scale scenes with line structured light.Firstly,the calibration of line structured light system is complicated.Secondly,among the method of reconstructing large-scale scenes using structured light,the shape based methods require rich texture information and complex geometric structures in the scene.The tracker based methods are not robust enough.The marker based methods are time-consuming and the laser based methods are not highly accurate.In order to solve the above problems,athree-dimensional scanner composed of binocular stereo vision systems and an infrared emitter is constructed in this paper.Among them,the binocular stereo vision system is used to reconstruct the projected infrared segment,thereby avoiding the complex calibration of line structured light system.At the same time,the high precision motion control system is used to control the uniform motion of the 3D scanner,and the reconstructed spatial lines are registered according to the motion parameters,which improves the accuracy and robustness of the method.Experimental results show that the proposed method is faster and more accurate than other methods.